“Are the contributions to GNOME and OpenOffice.org not enough?”
Clarification Openoffice.org is under LGPL and and a SUN specific license which also required every contributes to sign over copyright to SUN which means SUN is only vendor who can create a proprietary variant of Openoffice.org. Openoffice.org is NOT under GPL
As for GNOME, any derivative work of GNOME should be under GPL and contributors cant really play around with that
“Beside that, SUN also gave us NFS 🙂 ”
open specifications. not code
”
By using GRUB, they are not violating the GPL. If they find it necessary to change the software, then they ought to give back.
”
legally not but bad mouthing GPL extensively, supporting software patents in the full rush and then using it in software like GRUB an Samba is questionable ethics
“legally not but bad mouthing GPL extensively, supporting software patents in the full rush and then using it in software like GRUB an Samba is questionable ethics”
No, far from that. People may not like what SUN is doing, but that doesn’t mean they act unethical.
As for GNOME, any derivative work of GNOME should be under GPL and contributors cant really play around with that
No, despite all the pro free software handwaving by GNOME originally, GNOME is under a corporate-friendly LGPL license that helps them to produce closed source software. It is KDE that is under the GPL.
No, far from that. People may not like what SUN is doing, but that doesn’t mean they act unethical.
You’re right – it’s not unethical for Sun to do what they are doing, but it is additional evidence that the only relationship they want with the free software community is one where they take and don’t give back.
as long as it doesn’t break the GPL licence, even the fucking devil can use it – and people ought to STFU once and for all.
If you don’t like it, don’t release your software under the GPL but don’t MOAN and WHINE if you do, and then the code is used respecting the GPL to the letter.
OK. So, stop using GNOME, Mozilla, OpenOffice.org, Java, Ant, GNU lpr, Tomcat, Xalan, … and software that bundles those software/libraries
You’re in danger of occasionally touching some of a code/invention from an evil company.
Wash your hand throughoutly, using antibiotic soap*. And see a doctor for an advice immediately.
* Warning: some GPL/GPL-compatible SOAP implementations are in Java form, or do use Apache XML libraries (which that evil company do contribute to a lot).
“You’re right – it’s not unethical for Sun to do what they are doing, but it is additional evidence that the only relationship they want with the free software community is one where they take and don’t give back.”
Are you talking about the same company?
Sure, I have my own gripes about Sun and wonder what the … they are doing at times, though as others have pointed out they do indeed give back — quite a bit in fact.
If you want to complain about any company, complain about Microsoft or Apple. They both exhibit the behavior you attribute wrongly to Sun in spades.
RE: mvp (IP: —.hsd1.ca.comcast.net) – Posted on 2005-05-23 10:02:09
I wonder if it supports GPT (guid partition table).
That is an issue with the BIOS, GPT partitions can only be supported by EFI; now, there is an EFI reference available from Intel, too bad BIOS producers are too lazy to implement it 🙁
l3v1 (IP: —.vein.hu) – Posted on 2005-05-23 10:25:25
Solaris x86 will use GRUB as its boot manager on x86/x64 platforms
Uhmm, first question that pops: soooooowhat ?
That question can be answered if you took the time to look at Caspers post in reference to the benefits of using GRUB over the old system.
RE: PIssers and moaners
The only people complaining about SUN and GPL are those who are peeved that SUN didn’t grant GPL/LGPL programmes access to those 5000 patents. On one hand you have ‘code donated by IBM’, which is nothing more than road kill, and on the other hand, all you have to do to qualify to use the patents is use SUN’s new opensource license which is merely a variation on the old Mozilla Public License (MPL).
All features that would have appeared without IBMs help. How about IBM addressing the *REAL* issues that plague Linux, like the crap-o-la driver API and the lack of a stable one at that. How about the lack of applications on the workstation side of the equation. The lack of universal HIG that KDE and GNOME can conform to.
Sorry, I’ve said it before, and I’ll say it again; Linux’s weakness has NEVER been addressing the technical issues or the server problems (even Linus said that addressing the server issues are alot easier, as there are fixed requirements that need to be solved vs. the numerous issues that plague the desktop).
As for JFS; unneeded and only provided as to allow existing customers of AIX (which could fill a telephone booth, with enough room for a BBQ and a disco ball); EVMS was shunted out again in favour of LVMS.
All features that would have appeared without IBMs help.
Uh, no. Many were/are patented and/or IBM trade secrets, and all are crucial reasons why Linux has exploded in the enterprise marketplace. Your whining aside, IBM’s good reputation in the FOSS community is well deserved – as is Sun’s bad one. IBM joined the existing community and shared with it; Sun is trying to start a separate gated community and only take from the outside.
“as long as it doesn’t break the GPL licence, even the fucking devil can use it – and people ought to STFU once and for all. ”
Dont even bother indulging in censorship of opinion. Sure. What SUN is doing in legally right but morally wrong.
SUN’s dual faced approach to open source still continues despite their contributions. SUN engineers work their ass off in a good way and SUN management has this freaking attitude about pissing off on the GPL license, support for patents, lies about IBM, HP, Red Hat and what not
“How about IBM addressing the *REAL* issues that plague Linux, like the crap-o-la driver API and the lack of a stable one at that. ”
Linux user space like GTK is shared by SUN too. For the Linux kernel, their internal kernel API can changed anytime the developers want to. Read this
The only people complaining about SUN and GPL are those who are peeved that SUN didn’t grant GPL/LGPL programmes access to those 5000 patents. On one hand you have ‘code donated by IBM’, which is nothing more than road kill, and on the other hand, all you have to do to qualify to use the patents is use SUN’s new opensource license which is merely a variation on the old Mozilla Public License (MPL).”
IBM contributions are really to open source in the sense that they can be used by all OSI licenses but SUN made it available only to CDDL and claimed that it made it available to “open source”. read it at opensolaris.org (sic).
Now if they did donate it only to themselves via CDDL that is not a problem. just the postering about it
when it comes to licenses “mere” variations can mean a lot. LGPL is also a mere variant of the GPL but means quite different things. so the fact that CDDL itself is based on the MPL doesnt make it automatically acceptable. CDDL have patents clauses which only benefit SUN and not other redistributions possible when there exists opensolaris for the mass if and when it does. Also MPL is a failed disbalanced license. This is the reason why Mozilla foundation has trilicensed their code under GPL/LGPL/MPL.
Read OSI analysis of licenses and failure of MPL here
“All features that would have appeared without IBMs help.”
Uh, no. Many were/are patented and/or IBM trade secrets, and all are crucial reasons why Linux has exploded in the enterprise marketplace. Your whining aside, IBM’s good reputation in the FOSS community is well deserved – as is Sun’s bad one. IBM joined the existing community and shared with it; Sun is trying to start a separate gated community and only take from the outside.
Firstly, lets take NGPL, threading has been a known issue on Linux; why did IBM insist on working on a project *SEPERATELY* from the rest of the community, then suddenly expect it to be instantly merged into the kernel because it just so happened to be developed by big blue.
Secondly, why was Eclipsed released under the “Eclipse Public Licence”? I mean, you’re screaming about how SUN is using the CPPL to drive a wedge in the opensource community, and yet, IBM does the same thing through the development of *that* particular licence.
Thirdly, what is the issue with CPPL? have you actually thought that there *MAYBE* a reason why the deliberately *CHOSE* that licence? how about looking at the grander picture, SUNs eventual move to opensource their whole Java Enterprise Environment under CPPL, and how the Microsoft agreement fits into the larger picture.
May I suggest that you act like and adult, rather than taking the immature RMS of ‘all or nothing’ – how about looking for the grey middle ground; the compromise that allows the best of both worlds, and enables the end users to benefit – because ultimately *THEY ARE* the users of these products, and it will be the end users who will run Microsoft and the likes out of business, or atleast get them to change their business model. Work *IN* the system to make the change rather than simply being the roudy rabble rouser on the outside.
“What SUN is doing in legally right but morally wrong.”
I don’t think you understand the meaning of ‘moral’. I don’t even think you understand that Sun is a company. It is not their task (also in a moral way) to make YOUR life better.
“How about IBM addressing the *REAL* issues that plague Linux, like the crap-o-la driver API and the lack of a stable one at that.”
One of the weaknesses of Linux today but an advantage tomorrow is one of your *REAL* problems. If companies opensourced their drivers and provided references for future equipment, they will never have to bother with the maintainance of drivers anymore. The Linux people will write and maintain them.
I prefer it this way. If the Linux kernel hackers need to change direction with technology, then they can without being held up by having to be compatible with some 3rd party’s binary driver. This is one of the main reasons Linux has been able to progress so fast has been this freedom.
What exactly is Sun ‘taking’? They don’t *need* GRUB. That they choose to use it instead of their own yet-another-BM is Good. Ditto for Apple & KHTML. ”
SUN is taking GRUB and loads of other GPL software will their management continues to trash talk GPL in every “open solaris” meet as their reason for coming up with CDDL
That’s one big advantage that Linux has from an evaluation standpoint — it boots from a logical drive in the extended partition. Since Grub supports this, will Sun change their installation routine to allow this?
“I prefer it this way. If the Linux kernel hackers need to change direction with technology, then they can without being held up by having to be compatible with some 3rd party’s binary driver. This is one of the main reasons Linux has been able to progress so fast has been this freedom.”
This is one of the reason’s why Linux can’t progress to well onto the desktop – either big bloated kernel setups (even with modules) or incomplete hardware support, unless you compile yourself – which just takes to much times (and fails more and more often with code that needs to work together and haven’t been tested well enough – had enough of these before switching to freebsd for unix work).
And updating each driver every time another (minor) modification has been made, draws man power from other places in the kernel – not to mention that for real well working and accepted drivers, either the kernel hackers need to take care of, or the hw vendors need to join a good deal of kernel development, as they need to keep track of a good deal of changes, as we see drivers even fail between small version changes – if they are not updated.
And to what I heard some time ago on osnews, before it pops up again: No, always recompiling the kernel is NOT the reason linux might run more stable then windows. Windows’s bug “workaround”‘s at the kernel level and as it seems bad quality at basic services are.
I think that this is a wise decision. GRUB is a very good piece of software, and I think it deserves to be used and to be known more widely.
GRUB is an improvement over other software of that kind. I remember how much I hated to use LILO on Linux systems, because it always f*cked up the harddisk.
Solaris has always booted from an “extended” partition, it is of type 0x82 with multiple Solaris specific slices inside it. The behaviour that some Linux distros use of creating multiple fdisk paritions is against what the fdisk spec suggests. There is supposed to be one and only one partition of each type and it is supposed to be OS specific.
What changes with the new boot architecture (which replaces that old 16 bit real mode boot) is that a boot manager like grub can be used and Solaris lives much nicer with other systems on the same disk. It was always possible before but not easy.
It is also worth poining out that Solaris is multiboot complaint and can thus boot from any multiboot boot loader it doesn’t have to be grub. It happens to be grub because grub is good and is widely accepted.
At least it would make multibooting linux and solaris easier.
Although I’m not opposed to the use of GRUB, I’m not sure that it makes multibooting linux & solaris substantially better… by itself, anyway. Although I’m no fan of multibooting, I have my Ultra 5 set up to multiboot Debian Sarge & Solaris 10. Of course, SILO doesn’t actually boot Solaris, it invokes the Solaris loader instead. I can see that loading Solaris with GRUB would be faster than the SILO -> Solaris boot loader -> Solaris sequence, but my beef is that you can’t switch from Solaris to Linux without cycling power. If you try, Linux will load, but it will panic during initialization. Apparently Solaris does something to the hardware that Linux cannot, or at least does not, fix when it comes up. I would like to see this bit of unfriendliness eliminated.
Of course, I don’t, at present, find a compelling reason for booting Solaris, even Solaris 10, instead of Linux, nothwithstanding that I’ve devoted half the disk drive to it. I was thinking that having Solaris readily at hand would be a good thing, and it may yet be, but Linux still holds sway. I’m not a Debian devotee, but I’ve been pleasantly surprised by Sarge. Hmmm… I find I begin to stray off topic. Closing…
Solaris has always booted from an “extended” partition, it is of type 0x82 with multiple Solaris specific slices inside it.
You seem to be misusing x86 disk partitioning terminology.
Solaris can boot from a *primary* partition of the correct type (0x82) which contains its own filesystem slices, but it has not historically supported being installed or booted from a “logical drive” residing inside a DOS-style “extended partition” (type 0x05 or 0x0F).
Classic UNIX flavors such as FreeBSD and Solaris are almost alone in having this limitation these days, as most other non-Microsoft operating systems (OS/2, Linux, and BeOS for sure, and perhaps others) have no such limitation.
Perhaps it’s time for Sun to join the 1990’s? 🙂
The behaviour that some Linux distros use of creating multiple fdisk paritions is against what the fdisk spec suggests. There is supposed to be one and only one partition of each type and it is supposed to be OS specific.
Neither DOS, Windows, OS/2, BeOS, or Linux enforces this limit of one partition per disk of any given type.
If Solaris actually has such a limit, then it illustrates another articifial limitation in its ability to use disks properly on x86 hardware.
Such limitations are fine on Sparc hardware, but in the Intel world it has to compete for mindshare with other x86 operating systems. If I can’t install it where I already have free space, I’m not going to bother evaluating it…
That’s good! I like that (funny)! I realize that statements like mine inflame the Linux crowd, but forget the FUD, what are the benefits to using Linux over Solaris other than personal preference and “cool points”? Is it easier to administer, does it allow me the fine grained control that I might want to exert over user processes?
The issue I have with the more vocal Linux supporters (and I use that term loosely) that post here is that they are not concerned with technical merit, anyone who doesn’t use Linux has a problem and are wrong. If they don’t like Solaris, fine, so they don’t have to read the articles or post their FUD. If they want to be taken seriously, they should respond seriously.
Now maybe you want to answer the question (instead of dodging it) if you can. And why would I want to run Linux on the three (soon to be four) SunFire 4800’s (amongst other hardware) I have to administer (the vast majority of it SPARC)?
I know what I want to use, I prefer Solaris. Where I work we have Solaris (SPARC) and RedHat Linux (x86). I have used Linux off and on for years (since kernel 1.2.13) and Sun products since SunOS 4.1.3. What I want (and I’m sure others) to hear is something besides FUD from the Linux crowd, like actual technical reasons why I should look at Linux. If some of these people were trying to sell Linux to me based on what they are saying here, I would show them the door!
I just want to point out two things. First, part of the integration plan for GRUB by neccessity includes properly living up to the obligations of its license, which is the GPL:
$ pkginfo|grep -i grub
system SUNWgrub GNU GRUB – GNU GRand Unified Bootloader
system SUNWgrubS Source for GNU GRUB – GNU GRand Unified Bootloader
This means that source is provided. Our modifications to GRUB add a UFS filesystem reader. Those modifications are provided in source form. I am optimistic that these will wind up in the baseline GRUB distro.
The ‘copyright’ files for both packages include the text of the GPL. GRUB is a standalone program, and so my (non-lawyerly) understanding is that it is acceptable for GRUB to boot all operating systems, regardless of license.
Second, we are aware of the extended partition problem. It may not be our highest priority problem right now (for example, we’d rather get OpenSolaris out the door), but I’ll see if I can find the bugid and get it some attention.
Thanks for the feedback everyone. We appreciate it, as always.
Now maybe you want to answer the question (instead of dodging it) if you can.
What question is that?
And I should use Linux over Solaris (x86 or SPARC) because?
The 4800 series is Sparc architecture. It doesn’t make much sense to use Linux on Sparc. Why would you buy something like that if you want to run Linux, or even have a choice?
Its hard to answer stupid questions like why should you use Linux over Solaris on a SunFire 4800.
But on x86 the answer is obvious. Linux offers me more choice to use cheaper and sometimes obsolete hardware. I can provide the same services SunFire class hardware offers at a fraction of the price. And I can resell the OS, I have that choice as well.
Ever heard of clusters? They’re amazing. You see, you can take a bunch of PCs and bundle them together to meet or exceed the performance you get with a single-point-of-failure SunFire server. Sure you can cluster SunFires, too. But what’s the point? PCs give you more performance and more bang for your buck.
Cisco just replaced a big Sun enterprise build server with a cluster of PCs and some software from Trolltech and produced some impressive results. Its the results that matter, right? So why not try to save some money? That’s why I avoid Sun, unless I honestly need that class of hardware. Its a premium, not for average everyday workloads.
Tell me, how much did your SunFire setup cost you and what are you running on it that can’t be run on x86 hardware? How much x86 hardware would it take to provide the same services, if possible? And how much would that cost?
I really like it that SUN uses GRUB. It’s OK, it’s an excellent boot loader and SUN could use it even if there were no OpenSolaris at the horizon.
But there is one thing that I don’s like: SUN’s anti-GPL FUD. There are many articles on the web about SUN’s anti-GPL FUD. One point in SUN’s anti-GPL FUD is that GPL code cannot be combined with code under different licenses, which is wrong.
The fact that SUN uses GRUB for both OpenSolaris and ClosedSolaris proves that SUN’s anti-GPL FUD is wrong, GPL code can easily combined with code unter different licenses and now it’s finally proven.
SUN should use GRUB and in return stop its anti-GPL FUD, this would be the best solution for all parties.
Ah the standard Linux answer to “big iron” is clusters, not every application is cluster aware. Most of the applications (other than Oracle) that we use are not cluster aware. So the “cheap x86 cluster” argument doesn’t fly. And that “performacne” you talk about is also on specialized workloads, again not the every day stuff that big iron is used for.
It’s not just about saving money, there is the reliability issue. I don’t expect most PC servers to run as long or as reliably as a SPARC. And considering that these servers will probably be in place for 7 to 10 years, the places who buy this kind of hardware as a capitol investment.
If I wanted the ultimate in fault tolerance, I would have clustered SunFire machines using either Sun or Veritas Cluster software, I could (not recommended by Sun) split my 4800’s into 2 domains and using two machines, cluster the domains.
And what about IBM and HP, they sell similar hardware. So using your argument, customers who use IBM pSeries, zSeries and HP’s SuperDome and Integrity servers should replace them with PC’s running Linux? This hardware is just as expensive (if not more so) than Sun.
To provide the services we require (several large Oracle databases, document management, and a development environment) would probably run on PC servers, poorly. yes, it can be done with PC servers running Linux or Windows (Microsoft uses the same answer to address the shortcomings of PC hardware). Or I can use Solaris, AIX, or HP-UX on Enterprise grade hardware. Does it cost more, yes it does, is it more reliable, YES it is. And that is what people who spend the “big bucks” want.
I’ve worked with several clusters, the first one I setup was around 1998. One cluster option is to use a Localdirector or even a trick in DNS to load balance across a group of PCs. Another is to use software like Oracle or Trolltech’s Teambuilder or OpenMOSIX to provide load balancing, fault tolerance and redunancy, etc.
But if your decision to use “big iron” is based on your personal preference or trust in Sun hardware or some boost to your ego for spending the “big bucks” nothing I say will change your mind.
And, clusters do not offer a satisfactory solution for a very large class of softwares. Clusters are way overhyped, due in large part to PHBs not understanding Amdahl’s law, multithreading and the programmatic problem at hand. They just hear “many CPUs” and they are happy. Sadly Linux caters to such morons.
No, its about using the “right tool for the job”, and I don’t see that clustering a bunch of PC’s is better than big iron. Because if that was really true Sun, IBM, HP, SGI, NEC, Cray, and a host of others would be in big trouble. And since IBM and HP are porting Linux to their big iron hardware, it seems they are trying to “guild the lily” by providing Linux for their high-end stuff in order to give their potential customers a “choice”.
Also your comment about using obsolete hardware tells me you are willing to sacrifice performance since older hardware is not likely to perform as well as the newer stuff. I spent enough time when I was in the Navy using the “spit and bailing wire” method of keeping equipment up. I don’t have to anymore, and neither should a customer.
When reliability and proven (not theoretical) performance is more important, some customers prefer to spend the money for quality equipment and services. I work for customers who want to save money, but know what hardware and software they need to get the performance they want. And the savings comes as a result of getting the job done with minimum downtime. Not by buying cheap hardware and expecting it run like the “expensive stuff”.
No, its about using the “right tool for the job”, and I don’t see that clustering a bunch of PC’s is better than big iron. Because if that was really true Sun, IBM, HP, SGI, NEC, Cray, and a host of others would be in big trouble. And since IBM and HP are porting Linux to their big iron hardware, it seems they are trying to “guild the lily” by providing Linux for their high-end stuff in order to give their potential customers a “choice”.
Well, Sun’s financials aren’t exactly something to brag about mostly because they’re having a hard time selling the high margin big iron. Why do you think they now sell glorified PCs (aka their Opteron line)? SGI is damn near out of business and what’s left of them sells high end clustered super computers (which they like to sell with Linux installed on). Cray is big on clustered supercomputers as well these days. Given that you are trying to say that big iron is different from clustered supercomputers, I wouldn’t use SGI or Cray as an example. IBM is more of a services company then a hardware company but both HP and IBM will preinstall Linux on their big iron systems. Do you have any examples of companies that actually are doing well selling big iron systems that won’t install Linux on them for you?
Also your comment about using obsolete hardware tells me you are willing to sacrifice performance since older hardware is not likely to perform as well as the newer stuff. I spent enough time when I was in the Navy using the “spit and bailing wire” method of keeping equipment up. I don’t have to anymore, and neither should a customer.
Depending on the job I’d happily take obsolete hardware over expensive “big iron”. Who needs $15,000+ Sun systems for web heads when a $500 Dell throw away 1U does the job just fine? Even into the midrange, why pay big bucks for a Sun SPARC system when an IBM xSeries quad Xeon or whatever they call their new Opteron systems are faster and cheaper and have similar quality to the equivalent Sun systems?
When reliability and proven (not theoretical) performance is more important, some customers prefer to spend the money for quality equipment and services. I work for customers who want to save money, but know what hardware and software they need to get the performance they want. And the savings comes as a result of getting the job done with minimum downtime. Not by buying cheap hardware and expecting it run like the “expensive stuff”.
If someone is looking for reliability *and* performance, SPARC is the last place I’d send them. IA64 and Power both smoke the best that Sun or Fujitsu has available today. What is really sad is that an HP SuperDome running Win2k3 will handily outperform the best that Sun has to offer when it comes to stuff like databases. And if you need to absolute highest levels of I/O and reliability, you can’t beat a zSeries.
There really is a reason that Sun is having a hard time making a profit on their hardware, it just isn’t that good anymore.
The Integrity is HP’s way of trying to salvage their messed up strategy at the Enterprise level. Ask someone who is an existing HP customer if they want to get rid of their PA-RISC system to get Itanic, because that is what HP wants them to do. Wasn’t someone at HP not too long ago bitching about Sun “takin gaway their customers”? Yeah, people are lining up to buy the Integrity servers to run Windows, even you said it “when a $500 Dell throw away 1U does the job just fine”, the same people won’t spend that kind of money on an Integrity system because it’s not cheap!
One of the guys I work with at his previous place bought IBM xSeries, his view is they are junk! Constant replacing of hardware and minimal uptime. I have experience with IBM as well (AIX pSeries), it’s nice gear but not that nice. And have you priced those POWER 5 systems lately?
Yeah, some idiot bought us one of those Dell 2650’s, with no spare disks, no redundant power. Oh yeah, that’s what I want in my data center, NOT! Cheap is cheap, and I have heard this far too many times from PHB’s “ooh, look at this, we can save lots of money”. And the poor sysadmin has to figure out how to make it work.
As always the “financials” seem to come up, as if that is really a measure of system performance. Why don’t you save that for the “BSD is dying” troll?
Is the performance of the zSeries great because its a mainframe, or because you can run Linux in an LPAR? Unfortunately everything is NOT about Linux.
If you choose to buy whatever you want do so! For 12 years I have used Sun and intend to do so in the future. Just as the Linux crowd is loyal, so am I.
Somewhat odd that Sun would choose for a system running on enterprise-class machines a bootloader whose RAID support is problematic, to say the least.
– Robert Escue wrote:
Read the subject line, nowhere did it say SPARC. And I am sure that Sun is addressing the issues with grub and SVM (if there are any).
Umm, Robert, nowhere did *I* mention SPARC or SVM. I did mention “enterprise-class machines,” but there are certainly non-SPARC enterprise-class machines. Nor was I referring to SVM when mentioning RAID. I meant to include garden-variety hardware RAID, and I certainly meant to include machines with x86 processors.
Now instead of criticizing something I didn’t say, why don’t you try to install GRUB on any RAID system of your choosing, and report back to us on how entertaining the experience was.
Well considering that the V20z and V40z isn’t what I would call an “Enterprise Class” machines. Maybe what you need to do is be more specific, because we are using RedHat Enterprise Linux on our V20z’s (which use grub) and the onboard LSI RAID doesn’t seem to have any problems.
> If someone is looking for reliability *and* performance, SPARC is the last place I’d send them. IA64 and Power both smoke the best that Sun or Fujitsu has available today
There is a lot more to the servers than just CPU performance, plus it depends on how you define performance. If you shop for servers based on the processor benchmarks, you’re a fool, plain and simple. There are a lot more to be taken into consideration than some stupic SPECint/SPECfp benchmark. Yes, based on SPEC benchmarks SPARC is behind Itanic and Power, but at the same time SPARC is hell of a lot better value for the money in terms of investement protection and overall 64-bit story. Let me explain why.
With Itanic it is plain and obvious, the lack of binary compatibility, lack of good compilers and brain share already brought Itanic to canatonic state and I won’t be surprized if Itanic completely disappears in just a few years. Even HP lost its faith in IA64 and is now happily sells Opteron. IA64 SuperDome can’t really deliver and if you didn’t know 90% of all SuperDomes sold are still PA-RISC. I wonder what is going to happen to HP-UX when PA-RISC is completely EOL’ed (hint – dead).
With Power5 the story is not as simple but Power is still far from being perfect as lots of bigots claim it be.
* First off, raw performance wise the processor is pretty good, but even there IBM couldn’t play it straight. The SPEC benchmarks you see for Power are about 10 to 20 percent bullshit — the highest Power SPEC benchmarks are done on MCM’s (multi-core modules) and IBM used a dirty trick by turning off three out of four cores on the MCM giving all of the cache on the MCM to just one core, thus artificially improving the results. You can safely reduce the claimed SPEC results by about 10 to 20 percent to get the real world performance.
* Second, it looks like IBM didn’t get the memo about how much binary compatibility actually matters in this world — AIX on Power have been breaking binary compatibility with each generation of Power processor forcing major recompiles and box swaps. It looks like by IBM’s book the forklift upgrades are a day-to-day reality it seems (LOL). There is nothing like that with Sun — you can mix different family CPU’s in the same frame without any problems really extending the life your server and saving you $$$. Plus every break in binary compatibility reduces the number of applications available on the platform. Did you know that there are less than 200 certified application for Power5? You need to recompile to take advantage of Power5. Compare that with a few thousand of applications certified for SPARC (SPARC never broken binary compatibility in history).
* Third, micropartitioning on Power5 seriously sucks ass as it will sap almost 50% of CPU in just hypervisor overhead if you’ve got more that 10 micropartitions on the processor. With Solaris 10 zones there is no overhead whatsoever and a potential to run literally hundreds of virtual servers per CPU.
* Forth, AIX is grotesquely overpriced piece of garbage that is already so hopelessly behind Solaris that it is not even funny. Solaris is an absolute steal compared with AIX. IBM charges you about $1700 per CPU on averarge (depending on the box) and will charge you more for every LPAR or micropartition on top of that. Just to give you an example, AIX licensing on a 32 processor box comes just over $80,000 the last time I checked. Compare that with Solaris that can be had for free!
* Fifth, AIX on Power is still behind Solaris on SPARC in terms of RAS and DR (Dinamic Reconfiguration) in particular. Under AIX you still have to shutdown/reboot the entire box to perform maintenance on anything related to the CPU/memory, on Solaris/SPARC on the other hand you can replace uniboards on the live system without any problem whatsoever.
> why pay big bucks for a Sun SPARC system when an IBM xSeries quad Xeon or whatever they call their new Opteron systems are faster and cheaper and have similar quality to the equivalent Sun systems?
Err, UltraSparc powered Sun v440 with 4 processors is about the same price as 4 processor xSeries IBM crap. Only with Sun you buy the best proven and tested 64-bit platform, whereas with IBM you buy a glorified PC. Do you see a difference? Sun UltraSparc boxes are very competitive nowadays, regardless of whatever PC-minded morons tell you.
I’m sure the Linux crowd doesn’t want to hear what you just said, since they thrive on benchmarks regardless of whether they are based in reality or not. The POWER 4 based systems I worked on in a previous job weren’t all that fast, but they sure were expensive. I figured it would be pretty much the same with POWER 5.
HP is stoned if they think their high-end customers are going to pay to port their apps from PA-RISC to Itanium. And despite the fact that Linux runs on a SuperDome, I really can’t see a company spending a million dollars on a basically unproven platform to run Linux.
And like I have said before, if IBM and HP were “really” into helping Linux, they would “give to the community” the ability to create LPARs and vPars respectively to Linux. Don’t bank on that happening any time in the future. What’s the point in running on big iron if you can’t utilize big iron features.
> If someone is looking for reliability *and* performance, SPARC is the last place I’d send them. IA64 and Power both smoke the best that Sun or Fujitsu has available today
There is a lot more to the servers than just CPU performance, plus it depends on how you define performance. If you shop for servers based on the processor benchmarks, you’re a fool, plain and simple. There are a lot more to be taken into consideration than some stupic SPECint/SPECfp benchmark. Yes, based on SPEC benchmarks SPARC is behind Itanic and Power, but at the same time SPARC is hell of a lot better value for the money in terms of investement protection and overall 64-bit story. Let me explain why.
I never said there wasn’t more to servers than CPU performance, I said that if you want reliability and performance you don’t go with SPARC. None of your excuses changes that either, SPARC is slow. Perhaps you should spend less time defending a chip which even Sun can’t sell (see their Opteron boxes for more details) and start asking how they are going to save that sinking ship.
The Integrity is HP’s way of trying to salvage their messed up strategy at the Enterprise level. Ask someone who is an existing HP customer if they want to get rid of their PA-RISC system to get Itanic, because that is what HP wants them to do. Wasn’t someone at HP not too long ago bitching about Sun “takin gaway their customers”? Yeah, people are lining up to buy the Integrity servers to run Windows, even you said it “when a $500 Dell throw away 1U does the job just fine”, the same people won’t spend that kind of money on an Integrity system because it’s not cheap!
There are about the same number of people lined up to buy Sun’s big iron as HP’s (or IBM’s for that matter). Doesn’t change the fact that a SuperDome running Win2k3 would provide better performance for, for example, a database system then the comparable Sun system. Sun is getting so far behind in the performance game that it’s silly to even look at their gear if you aren’t already an all Sun shop.
One of the guys I work with at his previous place bought IBM xSeries, his view is they are junk! Constant replacing of hardware and minimal uptime. I have experience with IBM as well (AIX pSeries), it’s nice gear but not that nice. And have you priced those POWER 5 systems lately?
And I work on 2 xSeries and I’ve had no problems with either of them. 24×7 solid for over 2 years now. As for the price of a pSeries, you’re paying big bucks for some of the highest quality hardware you can buy. Isn’t that your whole argument, you pay more for better gear?
Yeah, some idiot bought us one of those Dell 2650’s, with no spare disks, no redundant power. Oh yeah, that’s what I want in my data center, NOT! Cheap is cheap, and I have heard this far too many times from PHB’s “ooh, look at this, we can save lots of money”. And the poor sysadmin has to figure out how to make it work.
Again, depends on what you need.
As always the “financials” seem to come up, as if that is really a measure of system performance. Why don’t you save that for the “BSD is dying” troll?
I didn’t bring financials up, you did when you said “Because if that was really true Sun, IBM, HP, SGI, NEC, Cray, and a host of others would be in big trouble.” Guess what, Sun, SGI and Cray are all in varying amounts of trouble because people aren’t buying the big iron. HP, IBM and NEC aren’t because they’re so diversified.
Is the performance of the zSeries great because its a mainframe, or because you can run Linux in an LPAR? Unfortunately everything is NOT about Linux.
Depends on what you mean by performance. They aren’t the fastest number crunchers out there. I said if you want I/O and reliability you buy a zSeries. (As for your LPAR comment, everything except z/OS and a handful of native apps run in an LPAR, that’s just how a zSeries works).
why pay big bucks for a Sun SPARC system when an IBM xSeries quad Xeon or whatever they call their new Opteron systems are faster and cheaper and have similar quality to the equivalent Sun systems?
Err, UltraSparc powered Sun v440 with 4 processors is about the same price as 4 processor xSeries IBM crap. Only with Sun you buy the best proven and tested 64-bit platform, whereas with IBM you buy a glorified PC. Do you see a difference? Sun UltraSparc boxes are very competitive nowadays, regardless of whatever PC-minded morons tell you.
Sun boxes might be priced competitively but you surely jest if you think they’re actually competetive in any other way. (Best proven and tested 64 bit platform? Who are you kidding. I just hope Sun chip designers stay away from the Opteron, it’d be sad if they manage to ruin two architectures)
And the IBM xSeries is hardly a glorified PC, if you’d ever used one you’d know that IBM uses all their own gear to make an enterprise level system using Intel processors. This in comparison to Sun’s Opteron line which are glorified PCs. They’re like Dells, basic whitebox gear in a pretty Sun case.
Sorry I’m offending all the Sun/SPARC lovers out there, but don’t shoot the messenger. Go ask Sun why their hardware performs so dismally. Go ask Sun why, if they think SPARC is so great, that they’ve had to release a line of Opteron systems (which, btw, is going to kill off the last vestigages of their low and midrange SPARC systems).
> I never said there wasn’t more to servers than CPU performance, I said that if you want reliability and performance you don’t go with SPARC. None of your excuses changes that either, SPARC is slow.
Fist of all, reliability should be off your list entirely. As I pointed out above SPARC systems from both Sun and Fujitsu are actually quite a bit ahead of both IBM and HP in terms of RAS. In respect to performance, yes, UltraSparc currently has some issues in respect to per-core single thread performance, but Sun can already equalize the per-core disparity by stuffing more cores on die and stuffing more chips into the system than their competitors. Sun FirePlane based systems scale more linearly than either IBM or HP and have no data skew, so having many processors on the backplane have less of decreasing marginal returns effect on overall performance. If single core performance is really your thing, than ask for Fujitsu PrimePower with SPARC64, the 2Ghz chip has single thread performance pretty much on par with the latest Power5 offering. Plus with Sun and Fujistu joining their chip development operations into APL by the end of this year, the “SPARC is slow” garbage is going to become history anyway.
On the side note, besides everything said above, as they say “you date hardware, but you marry your OS”. You can flirt with hardware all you want, but the real choice should come down to the operating system. And on that front Solaris 10 is the OS you would want to marry, not AIX or HP or Linux or god forbid Windows. Solaris 10 is head and sholders above the rest.
> There are about the same number of people lined up to buy Sun’s big iron as HP’s (or IBM’s for that matter). Doesn’t change the fact that a SuperDome running Win2k3 would provide better performance for, for example, a database system then the comparable Sun system. Sun is getting so far behind in the performance game that it’s silly to even look at their gear if you aren’t already an all Sun shop.
I’m not sure where you get these ideas that Sun somehow hopelessly behind HP and IBM in databases, I hope you didn’t drink too much of that IBM koolaid with whatever they put in it. Just as sort of a reality check consider this, 2 out 3 Oracle deployments still go on Solaris/Sparc. The biggest data warehouses in the world run on Solaris/Sparc. And SuperDome with Win2k3, are you kidding? This is just laughable! I bet HP haven’t sold a single one in that configuration. As far as database applications are concerned (especially highly transactional) it is not about CPU performance, it more about memory bandwidth and memory access symmetry. If you need more convincing why Sun systems are still the best for database applications and data wharehousing in general, may be this whitepaper from the creators of the biggest data wharehouses in the world can enlighten you:
> Go ask Sun why their hardware performs so dismally. Go ask Sun why, if they think SPARC is so great, that they’ve had to release a line of Opteron systems (which, btw, is going to kill off the last vestigages of their low and midrange SPARC systems).
If unit sales are of any indicator, SPARC systems are still very popular and aparently many customers still love them (yours truly included). According to the report from Sun, the unit shipments for SPARC gear increased by 20 odd percent in the last year alone. It doesn’t look it is whithering away, that’s for sure. And why do you thing the fact that Sun started selling Opteron has any inclination to SPARC? Sun is trying to grab as much as possible at the low end to improve market penetration. Plus, since Sun is not just a box dropper any more and sells software, storage and services, if they can attach a middleware or storage sale to some low end Opteron kit even if it ends up running Linux, Sun still wins. They try to diversify from just being a narrowly focused server vendor and this is a good thing.
Have you actually used a V20z or a V40z, or are you just talking trash? Just two feet away from me is three V20z’s (of the ten we purchased) and I hardly consider them PC’s. There are some things I don’t like about them, but I would rather have them than Dell boxes.
And just who is going to spend $250,000 (2003 dollars) on a minimally configured Intergrity SuperDome to run Windows?
No when it comes to IBM gear, it is just high priced. And I just love getting a “crazy 8’s” over how you plug a cord into a power supply!
It’s pretty simple. If they cannot sell their hardware then they will cease to exist. I don’t think you’re particularly qualified to discuss performance issues (at least from every single post I’ve read of yours, your ability to discuss technical performance issues seems to treat your conclusions as axiomatic and thus not particularly interesting), so I don’t really feel like being part of your conversation.
If there is such a tremendous demand for their hardware and what you think it does, then I suggest that you shore up your suppliers’ financials for them by purchasing it. You seem to think that these companies shouldn’t be worried, when their financial trends indicate that if they aren’t they’re not going to be around, or they’re going to have to cut expenses to reflect selling niche products. That’s basically what SGI has been doing. Who knows what HP and Sun are going to decide on doing; I doubt they do.
Well if you don’t like my discussions on performance, don’t particpate! I don’t mind (as I am sure others here won’t as well). Maybe you should read my posts at http://www.sysadmintalk.com, where system administrators help each other. Unlike the mostly FUD filled discussions here, at sysadmintalk we actually talk about REAL issues, like Oracle with and without DIRECTIO on disks (amongst other things). Unfortunately real performance topics are not discussed here.
If you want to talk about specific performance issues, fine. I think I’m more than capable, are you?
In his second and third responses to my following post –
Somewhat odd that Sun would choose for a system running on enterprise-class machines a bootloader whose RAID support is problematic, to say the least.
– Robert Escue said:
Well considering that the V20z and V40z isn’t what I would call an “Enterprise Class” machines. Maybe what you need to do is be more specific, because we are using RedHat Enterprise Linux on our V20z’s (which use grub) and the onboard LSI RAID doesn’t seem to have any problems.
I don’t exactly call IDE RAID “Enterprise Class” stuff, I prefer SCSI and Fibre Channel myself.
Your first response is to say that two machines I never mentioned aren’t “enterprise class.” Umm, fine, except it seems entirely irrelevant to the subject at hand. You also state that GRUB appears to run fine on some machines where you work that have LSI RAID. Did you set up GRUB on these machines? If so, please tell us the steps you took, and if you would characterize the process as easier or more difficult than average.
Your second response cites one instance of difficulty installing GRUB on IDE RAID, and two notes saying *the 2.6 Linux kernel* will no longer support certain software RAID cards. You then extrapolate from these to a remark that you don’t consider IDE “enterprise class.”
In comments responding to others in this thread you’ve praised your own technical competence. Thus I find it puzzling that you’ve apparently confused RAID support in the Linux kernel with problems in the GRUB software application related to RAID. It’s also disconcerting that you would, from a single notice of GRUB problems with IDE RAID, jump to the conclusion that this problem is restricted to IDE. It’s not. You can easily confirm this with some quick searches relating to GRUB, RAID, and SCSI.
Well, firstly let me say that I find sysadmintalk.com an excellent, very pertinent resource. Also, I have been working with HP, IBM and Sun hardware (and HP-UX, Linux and Solaris) for almost a decade now. Also, I have no reason to doubt Robert’s technical expertise, quite the contrary, I think he’s right on the money in what he says here.
However, said all that, I am not a big fan of GRUB. There are several things I dislike about it, and really didn’t feel confortable working with it.
Well, I am sure Sun has real expert making these kinds of decision – I am nowhere qualified enough to make judgement of any kind, about x86 Solaris bootloader.
Yes grub was set up on all of the V20z’s with no issues at all. Just install as you would on a machine without RAID. And our machines have survived multiple reboots with no loss of drives.
Now you want specifc answers, post specific questions instead of your blathering bullshit.
“it is additional evidence that the only relationship they want with the free software community is one where they take and don’t give back”
What about Open Office? NFS? RPC? Their contributions to Gnome? What do you think Open Solaris is all about? (and what about all the patents they donated?) Snu has made major contributions to OSS and these continuing claims that they haven’t are pure FUD.
I have no idea what they want, but obviously the contributions Sun has made is not enough. I am also trying to figure out the animosity that the Linux zealots have about Sun and Solaris. Most of the arguments I have read here are absolutely ridiculous like “I don’t like what Jonathan Schwartz has to say”. Like that really affects my decision making process as to whether or not to use a product.
I would just like to see an article or link that has anything to do with Solaris not become a FUD fest. You would think the Linux crowd would be more focused on trying to persuade others to the benefits of Linux online and off, rather than anger those they might have to spend more time convincing. Or at least that’s what I did when I was part of “Team OS/2”.
I think you missed the biggest issue; the latency issue on large POWER boxes when compared to SPARC. Now, sure, I’m all for bashing *UltraSPARC*, thats an easy target, but price/performance wise, SPARC64 isn’t half bad. IMHO SUN should replace the UltraSPARC III cores in their UltraSPARC IV with SPARC64, and release it as a UltraSPARC V, that alone, dual core, running at 2.1Ghz per core would provide a damn good power boost to their enterprise line up.
Oh, and just as a side issue, SPARC is an openstandard. Unlike POWER which is controlled by IBM – oh, and don’t believe the OpenPOWER hype – they’ve only released a *SMALL* subset of the ISA to the market, and even then, there are BIG strings attached to it.
“I think you missed the biggest issue; the latency issue on large POWER boxes when compared to SPARC. Now, sure, I’m all for bashing *UltraSPARC*, thats an easy target, but price/performance wise, SPARC64 isn’t half bad. IMHO SUN should replace the UltraSPARC III cores in their UltraSPARC IV with SPARC64, and release it as a UltraSPARC V, that alone, dual core, running at 2.1Ghz per core would provide a damn good power boost to their enterprise line up.”
In response to my observation that I found it odd Sun would adopt GRUB as its bootloader for enterprise-class machines in view of some problems GRUB had with RAID, Robert Escue’s latest reply was:
Yes grub was set up on all of the V20z’s with no issues at all. Just install as you would on a machine without RAID. And our machines have survived multiple reboots with no loss of drives.
Now you want specific answers, post specific questions instead of your blathering bull***t.
I don’t recall blathering, nor asking questions, simply making an observation. Regarding specifics, GRUB has had and continues AFAIK to have problems identifying, accessing, and being installed to IDE or SCSI disks in RAID arrays (other than RAID1, where installation at least in certain Linux distros is now rather routine).
A decent non-technical explanation appears in this March 2005 e-mail to the bug-grub mailing list:
I am also trying to figure out the animosity that the Linux zealots have about Sun and Solaris. Most of the arguments I have read here are absolutely ridiculous like [etc…]
Of course, if weren’t for Linux zealots, all would be peace and harmony, wouldn’t it? NOT! Maybe if there were no zealots, perhaps. Every advocacy group has zealots and they are usually rude, stupid, fatuous and self-serving. Sun does not help the matter with its own brand of FUD.
I would just like to see an article or link that has anything to do with Solaris not become a FUD fest.
That would be remarkable, wouldn’t it?
You would think the Linux crowd would be more focused on trying to persuade others to the benefits of Linux online and off…
And you would think that “the Solaris crowd” would be more focused on trying to persuade others of the benefits of Solaris online and off rather than telling me how bad Linux is, how inferior it is and, oh, by the way, how naughty Red Hat is. Solaris and its advocates would be better served if they would busy themselves with showing why Solaris is good and to leave off trying to tell me to hate Linux. Failing that, they can at least shut up, along with all other zealots, of course.
… rather than anger those they might have to spend more time convincing. Or at least that’s what I did when I was part of “Team OS/2”.
Maybe you did. However, I, too, was a part of Team OS/2, and was well acquainted with the good, the bad and ugly of that team effort. I would like to think that the nebulous group which umbrella’d under the appellation “Team OS/2” was better than most advocacy groups, but it wasn’t, not on average. As usual, only individuals stood out as measured and reasonable, whereas others were merely loud and vituperative, just like every other advocacy group.
We are often less critical of those whose views are more congenial with our own and much more so of those who disagree. After all, since we’re right and they’re wrong, we are only speaking in the tones of righteousness whereas they are being shrill and abusive. It was ever thus.
Harking back to the days of Windows vs. OS/2 is both amusing and pathetic since you (presumably) liken Linux to Windows (the bad guys) and Solaris to OS/2 (the good guys), when not so long ago, the face-off was reckoned to be the Windows bad guys vs. the Linux good guys.
> Are the contributions to GNOME and OpenOffice.org not enough?
Beside that, SUN also gave us NFS 🙂
Carsten
From you reaction it seems you are not completely happy with the rules of the GPL.
By using GRUB, they are not violating the GPL. If they find it necessary to change the software, then they ought to give back.
I wonder if it supports GPT (guid partition table).
Any comparisons available?
Carsten
“Are the contributions to GNOME and OpenOffice.org not enough?”
Clarification Openoffice.org is under LGPL and and a SUN specific license which also required every contributes to sign over copyright to SUN which means SUN is only vendor who can create a proprietary variant of Openoffice.org. Openoffice.org is NOT under GPL
As for GNOME, any derivative work of GNOME should be under GPL and contributors cant really play around with that
“Beside that, SUN also gave us NFS 🙂 ”
open specifications. not code
”
By using GRUB, they are not violating the GPL. If they find it necessary to change the software, then they ought to give back.
”
legally not but bad mouthing GPL extensively, supporting software patents in the full rush and then using it in software like GRUB an Samba is questionable ethics
“legally not but bad mouthing GPL extensively, supporting software patents in the full rush and then using it in software like GRUB an Samba is questionable ethics”
No, far from that. People may not like what SUN is doing, but that doesn’t mean they act unethical.
Mmm. Possibly… But even companies with questionable ethics are allowed to use GPL software – providing that is they don’t break the rules of the GPL.
GJ
Solaris x86 will use GRUB as its boot manager on x86/x64 platforms
Uhmm, first question that pops: soooooowhat ?
As for GNOME, any derivative work of GNOME should be under GPL and contributors cant really play around with that
No, despite all the pro free software handwaving by GNOME originally, GNOME is under a corporate-friendly LGPL license that helps them to produce closed source software. It is KDE that is under the GPL.
Does Sun granting the use of 5000 patents for OSS use counting as something valuable?
Sun is probably the most OSS friendly corporation out there.
No, far from that. People may not like what SUN is doing, but that doesn’t mean they act unethical.
You’re right – it’s not unethical for Sun to do what they are doing, but it is additional evidence that the only relationship they want with the free software community is one where they take and don’t give back.
“Solaris x86 will use GRUB as its boot manager on x86/x64 platforms
Uhmm, first question that pops: soooooowhat ? ”
Yeah !!! Sooowhat !!! It’s free !!!
as long as it doesn’t break the GPL licence, even the fucking devil can use it – and people ought to STFU once and for all.
If you don’t like it, don’t release your software under the GPL but don’t MOAN and WHINE if you do, and then the code is used respecting the GPL to the letter.
OK. So, stop using GNOME, Mozilla, OpenOffice.org, Java, Ant, GNU lpr, Tomcat, Xalan, … and software that bundles those software/libraries
You’re in danger of occasionally touching some of a code/invention from an evil company.
Wash your hand throughoutly, using antibiotic soap*. And see a doctor for an advice immediately.
* Warning: some GPL/GPL-compatible SOAP implementations are in Java form, or do use Apache XML libraries (which that evil company do contribute to a lot).
sincerely yours,
Nokia? Nokia? Wasn’t there something about Nokia pressing for Software Patents in Europe as well?
“You’re right – it’s not unethical for Sun to do what they are doing, but it is additional evidence that the only relationship they want with the free software community is one where they take and don’t give back.”
Are you talking about the same company?
Sure, I have my own gripes about Sun and wonder what the … they are doing at times, though as others have pointed out they do indeed give back — quite a bit in fact.
If you want to complain about any company, complain about Microsoft or Apple. They both exhibit the behavior you attribute wrongly to Sun in spades.
RE: mvp (IP: —.hsd1.ca.comcast.net) – Posted on 2005-05-23 10:02:09
I wonder if it supports GPT (guid partition table).
That is an issue with the BIOS, GPT partitions can only be supported by EFI; now, there is an EFI reference available from Intel, too bad BIOS producers are too lazy to implement it 🙁
l3v1 (IP: —.vein.hu) – Posted on 2005-05-23 10:25:25
Solaris x86 will use GRUB as its boot manager on x86/x64 platforms
Uhmm, first question that pops: soooooowhat ?
That question can be answered if you took the time to look at Caspers post in reference to the benefits of using GRUB over the old system.
RE: PIssers and moaners
The only people complaining about SUN and GPL are those who are peeved that SUN didn’t grant GPL/LGPL programmes access to those 5000 patents. On one hand you have ‘code donated by IBM’, which is nothing more than road kill, and on the other hand, all you have to do to qualify to use the patents is use SUN’s new opensource license which is merely a variation on the old Mozilla Public License (MPL).
On one hand you have ‘code donated by IBM’, which is nothing more than road kill
LOL. Just to the Linux kernel here are a few things that outweigh Sun’s whole list:
Read Copy Update (RCU)
Journaling File System (JFS)
The Enterprise Volume Management System (EVMS)
Asynchronous input-output operations (AIO)
Direct input-output operations (DIO)
Non-uniform memory access (NUMA)
Symmetric multi-processor support (SMP)
All features that would have appeared without IBMs help. How about IBM addressing the *REAL* issues that plague Linux, like the crap-o-la driver API and the lack of a stable one at that. How about the lack of applications on the workstation side of the equation. The lack of universal HIG that KDE and GNOME can conform to.
Sorry, I’ve said it before, and I’ll say it again; Linux’s weakness has NEVER been addressing the technical issues or the server problems (even Linus said that addressing the server issues are alot easier, as there are fixed requirements that need to be solved vs. the numerous issues that plague the desktop).
As for JFS; unneeded and only provided as to allow existing customers of AIX (which could fill a telephone booth, with enough room for a BBQ and a disco ball); EVMS was shunted out again in favour of LVMS.
Sun granted 5000 patents ONLY to CDDL developers. If you use the most popular license by an order of maganitude (GPL) you can still get sued by Sun.
IBM has granted their patents to Everybody under ANY OSI approved License including the CDDL.
Sun’s developers know how to use the GPL, and be friendly to OSS. Sun’s management still want to be like MSFT.
All features that would have appeared without IBMs help.
Uh, no. Many were/are patented and/or IBM trade secrets, and all are crucial reasons why Linux has exploded in the enterprise marketplace. Your whining aside, IBM’s good reputation in the FOSS community is well deserved – as is Sun’s bad one. IBM joined the existing community and shared with it; Sun is trying to start a separate gated community and only take from the outside.
At least it would make multibooting linux and solaris easier.
“as long as it doesn’t break the GPL licence, even the fucking devil can use it – and people ought to STFU once and for all. ”
Dont even bother indulging in censorship of opinion. Sure. What SUN is doing in legally right but morally wrong.
SUN’s dual faced approach to open source still continues despite their contributions. SUN engineers work their ass off in a good way and SUN management has this freaking attitude about pissing off on the GPL license, support for patents, lies about IBM, HP, Red Hat and what not
“How about IBM addressing the *REAL* issues that plague Linux, like the crap-o-la driver API and the lack of a stable one at that. ”
Linux user space like GTK is shared by SUN too. For the Linux kernel, their internal kernel API can changed anytime the developers want to. Read this
http://www.kroah.com/log/2004/12/03/#stable_api_nonsense
The screenshots didn’t load.Anyone else experiencing this?
”
The only people complaining about SUN and GPL are those who are peeved that SUN didn’t grant GPL/LGPL programmes access to those 5000 patents. On one hand you have ‘code donated by IBM’, which is nothing more than road kill, and on the other hand, all you have to do to qualify to use the patents is use SUN’s new opensource license which is merely a variation on the old Mozilla Public License (MPL).”
IBM contributions are really to open source in the sense that they can be used by all OSI licenses but SUN made it available only to CDDL and claimed that it made it available to “open source”. read it at opensolaris.org (sic).
Now if they did donate it only to themselves via CDDL that is not a problem. just the postering about it
when it comes to licenses “mere” variations can mean a lot. LGPL is also a mere variant of the GPL but means quite different things. so the fact that CDDL itself is based on the MPL doesnt make it automatically acceptable. CDDL have patents clauses which only benefit SUN and not other redistributions possible when there exists opensolaris for the mass if and when it does. Also MPL is a failed disbalanced license. This is the reason why Mozilla foundation has trilicensed their code under GPL/LGPL/MPL.
Read OSI analysis of licenses and failure of MPL here
http://opensource.org/docs/policy/licenseproliferation.php
“All features that would have appeared without IBMs help.”
Uh, no. Many were/are patented and/or IBM trade secrets, and all are crucial reasons why Linux has exploded in the enterprise marketplace. Your whining aside, IBM’s good reputation in the FOSS community is well deserved – as is Sun’s bad one. IBM joined the existing community and shared with it; Sun is trying to start a separate gated community and only take from the outside.
Firstly, lets take NGPL, threading has been a known issue on Linux; why did IBM insist on working on a project *SEPERATELY* from the rest of the community, then suddenly expect it to be instantly merged into the kernel because it just so happened to be developed by big blue.
Secondly, why was Eclipsed released under the “Eclipse Public Licence”? I mean, you’re screaming about how SUN is using the CPPL to drive a wedge in the opensource community, and yet, IBM does the same thing through the development of *that* particular licence.
Thirdly, what is the issue with CPPL? have you actually thought that there *MAYBE* a reason why the deliberately *CHOSE* that licence? how about looking at the grander picture, SUNs eventual move to opensource their whole Java Enterprise Environment under CPPL, and how the Microsoft agreement fits into the larger picture.
May I suggest that you act like and adult, rather than taking the immature RMS of ‘all or nothing’ – how about looking for the grey middle ground; the compromise that allows the best of both worlds, and enables the end users to benefit – because ultimately *THEY ARE* the users of these products, and it will be the end users who will run Microsoft and the likes out of business, or atleast get them to change their business model. Work *IN* the system to make the change rather than simply being the roudy rabble rouser on the outside.
should be CDDL not CPPL (yes, I feel like a right nit-witt)
“What SUN is doing in legally right but morally wrong.”
I don’t think you understand the meaning of ‘moral’. I don’t even think you understand that Sun is a company. It is not their task (also in a moral way) to make YOUR life better.
“How about IBM addressing the *REAL* issues that plague Linux, like the crap-o-la driver API and the lack of a stable one at that.”
One of the weaknesses of Linux today but an advantage tomorrow is one of your *REAL* problems. If companies opensourced their drivers and provided references for future equipment, they will never have to bother with the maintainance of drivers anymore. The Linux people will write and maintain them.
I prefer it this way. If the Linux kernel hackers need to change direction with technology, then they can without being held up by having to be compatible with some 3rd party’s binary driver. This is one of the main reasons Linux has been able to progress so fast has been this freedom.
sun should not use gpl software while not beeing compatible to gpl. they only take but don’t give back…
What exactly is Sun ‘taking’? They don’t *need* GRUB. That they choose to use it instead of their own yet-another-BM is Good. Ditto for Apple & KHTML.
What have you, I assume a user of GPL’d software and thence the one that really benefits from it, have ever ‘given back’??
”
What exactly is Sun ‘taking’? They don’t *need* GRUB. That they choose to use it instead of their own yet-another-BM is Good. Ditto for Apple & KHTML. ”
SUN is taking GRUB and loads of other GPL software will their management continues to trash talk GPL in every “open solaris” meet as their reason for coming up with CDDL
Greetings,
Just to be “On-topic”, I would like to say that the OpenSolaris port to PPC (also referred as Polaris) will use GNU GRUB2.
Regards,
Jay
That’s one big advantage that Linux has from an evaluation standpoint — it boots from a logical drive in the extended partition. Since Grub supports this, will Sun change their installation routine to allow this?
I’d like to add that the OpenSolatis port to PPC will be able to factor arbitrary integers in polynomial time.
Thank I was not aware it was only their Open Source license. Still thats not too bad.
Correct me if I am wrong but of the 500 patents IBM released most are worthless and some are not even relevant to software.
I think IBM statement that they will use their patents to defend the Linux kernel is more important than their patent grants.
“I prefer it this way. If the Linux kernel hackers need to change direction with technology, then they can without being held up by having to be compatible with some 3rd party’s binary driver. This is one of the main reasons Linux has been able to progress so fast has been this freedom.”
This is one of the reason’s why Linux can’t progress to well onto the desktop – either big bloated kernel setups (even with modules) or incomplete hardware support, unless you compile yourself – which just takes to much times (and fails more and more often with code that needs to work together and haven’t been tested well enough – had enough of these before switching to freebsd for unix work).
And updating each driver every time another (minor) modification has been made, draws man power from other places in the kernel – not to mention that for real well working and accepted drivers, either the kernel hackers need to take care of, or the hw vendors need to join a good deal of kernel development, as they need to keep track of a good deal of changes, as we see drivers even fail between small version changes – if they are not updated.
And to what I heard some time ago on osnews, before it pops up again: No, always recompiling the kernel is NOT the reason linux might run more stable then windows. Windows’s bug “workaround”‘s at the kernel level and as it seems bad quality at basic services are.
I think that this is a wise decision. GRUB is a very good piece of software, and I think it deserves to be used and to be known more widely.
GRUB is an improvement over other software of that kind. I remember how much I hated to use LILO on Linux systems, because it always f*cked up the harddisk.
They’re gonna need a lot more than GRUB to save Solaris X86.
Solaris has always booted from an “extended” partition, it is of type 0x82 with multiple Solaris specific slices inside it. The behaviour that some Linux distros use of creating multiple fdisk paritions is against what the fdisk spec suggests. There is supposed to be one and only one partition of each type and it is supposed to be OS specific.
What changes with the new boot architecture (which replaces that old 16 bit real mode boot) is that a boot manager like grub can be used and Solaris lives much nicer with other systems on the same disk. It was always possible before but not easy.
It is also worth poining out that Solaris is multiboot complaint and can thus boot from any multiboot boot loader it doesn’t have to be grub. It happens to be grub because grub is good and is widely accepted.
And I should use Linux over Solaris (x86 or SPARC) because?
At least it would make multibooting linux and solaris easier.
Although I’m not opposed to the use of GRUB, I’m not sure that it makes multibooting linux & solaris substantially better… by itself, anyway. Although I’m no fan of multibooting, I have my Ultra 5 set up to multiboot Debian Sarge & Solaris 10. Of course, SILO doesn’t actually boot Solaris, it invokes the Solaris loader instead. I can see that loading Solaris with GRUB would be faster than the SILO -> Solaris boot loader -> Solaris sequence, but my beef is that you can’t switch from Solaris to Linux without cycling power. If you try, Linux will load, but it will panic during initialization. Apparently Solaris does something to the hardware that Linux cannot, or at least does not, fix when it comes up. I would like to see this bit of unfriendliness eliminated.
Of course, I don’t, at present, find a compelling reason for booting Solaris, even Solaris 10, instead of Linux, nothwithstanding that I’ve devoted half the disk drive to it. I was thinking that having Solaris readily at hand would be a good thing, and it may yet be, but Linux still holds sway. I’m not a Debian devotee, but I’ve been pleasantly surprised by Sarge. Hmmm… I find I begin to stray off topic. Closing…
you guys need to read this post;
about WHY the change to grub happend… makes a LOT of sense to me…..
http://blogs.sun.com/roller/page/casper/20050427#the_end_of_realmod…
Solaris has always booted from an “extended” partition, it is of type 0x82 with multiple Solaris specific slices inside it.
You seem to be misusing x86 disk partitioning terminology.
Solaris can boot from a *primary* partition of the correct type (0x82) which contains its own filesystem slices, but it has not historically supported being installed or booted from a “logical drive” residing inside a DOS-style “extended partition” (type 0x05 or 0x0F).
Classic UNIX flavors such as FreeBSD and Solaris are almost alone in having this limitation these days, as most other non-Microsoft operating systems (OS/2, Linux, and BeOS for sure, and perhaps others) have no such limitation.
Perhaps it’s time for Sun to join the 1990’s? 🙂
The behaviour that some Linux distros use of creating multiple fdisk paritions is against what the fdisk spec suggests. There is supposed to be one and only one partition of each type and it is supposed to be OS specific.
Neither DOS, Windows, OS/2, BeOS, or Linux enforces this limit of one partition per disk of any given type.
If Solaris actually has such a limit, then it illustrates another articifial limitation in its ability to use disks properly on x86 hardware.
Such limitations are fine on Sparc hardware, but in the Intel world it has to compete for mindshare with other x86 operating systems. If I can’t install it where I already have free space, I’m not going to bother evaluating it…
I’m sorry, did I mention Linux somewhere in my post?
Perhaps it was the voices in your head.. I’d listen to them if I were you. Smart voices.
That’s good! I like that (funny)! I realize that statements like mine inflame the Linux crowd, but forget the FUD, what are the benefits to using Linux over Solaris other than personal preference and “cool points”? Is it easier to administer, does it allow me the fine grained control that I might want to exert over user processes?
The issue I have with the more vocal Linux supporters (and I use that term loosely) that post here is that they are not concerned with technical merit, anyone who doesn’t use Linux has a problem and are wrong. If they don’t like Solaris, fine, so they don’t have to read the articles or post their FUD. If they want to be taken seriously, they should respond seriously.
Now maybe you want to answer the question (instead of dodging it) if you can. And why would I want to run Linux on the three (soon to be four) SunFire 4800’s (amongst other hardware) I have to administer (the vast majority of it SPARC)?
I know what I want to use, I prefer Solaris. Where I work we have Solaris (SPARC) and RedHat Linux (x86). I have used Linux off and on for years (since kernel 1.2.13) and Sun products since SunOS 4.1.3. What I want (and I’m sure others) to hear is something besides FUD from the Linux crowd, like actual technical reasons why I should look at Linux. If some of these people were trying to sell Linux to me based on what they are saying here, I would show them the door!
I just want to point out two things. First, part of the integration plan for GRUB by neccessity includes properly living up to the obligations of its license, which is the GPL:
$ pkginfo|grep -i grub
system SUNWgrub GNU GRUB – GNU GRand Unified Bootloader
system SUNWgrubS Source for GNU GRUB – GNU GRand Unified Bootloader
This means that source is provided. Our modifications to GRUB add a UFS filesystem reader. Those modifications are provided in source form. I am optimistic that these will wind up in the baseline GRUB distro.
The ‘copyright’ files for both packages include the text of the GPL. GRUB is a standalone program, and so my (non-lawyerly) understanding is that it is acceptable for GRUB to boot all operating systems, regardless of license.
Second, we are aware of the extended partition problem. It may not be our highest priority problem right now (for example, we’d rather get OpenSolaris out the door), but I’ll see if I can find the bugid and get it some attention.
Thanks for the feedback everyone. We appreciate it, as always.
Now maybe you want to answer the question (instead of dodging it) if you can.
What question is that?
And I should use Linux over Solaris (x86 or SPARC) because?
The 4800 series is Sparc architecture. It doesn’t make much sense to use Linux on Sparc. Why would you buy something like that if you want to run Linux, or even have a choice?
Its hard to answer stupid questions like why should you use Linux over Solaris on a SunFire 4800.
But on x86 the answer is obvious. Linux offers me more choice to use cheaper and sometimes obsolete hardware. I can provide the same services SunFire class hardware offers at a fraction of the price. And I can resell the OS, I have that choice as well.
Ever heard of clusters? They’re amazing. You see, you can take a bunch of PCs and bundle them together to meet or exceed the performance you get with a single-point-of-failure SunFire server. Sure you can cluster SunFires, too. But what’s the point? PCs give you more performance and more bang for your buck.
Cisco just replaced a big Sun enterprise build server with a cluster of PCs and some software from Trolltech and produced some impressive results. Its the results that matter, right? So why not try to save some money? That’s why I avoid Sun, unless I honestly need that class of hardware. Its a premium, not for average everyday workloads.
Tell me, how much did your SunFire setup cost you and what are you running on it that can’t be run on x86 hardware? How much x86 hardware would it take to provide the same services, if possible? And how much would that cost?
I really like it that SUN uses GRUB. It’s OK, it’s an excellent boot loader and SUN could use it even if there were no OpenSolaris at the horizon.
But there is one thing that I don’s like: SUN’s anti-GPL FUD. There are many articles on the web about SUN’s anti-GPL FUD. One point in SUN’s anti-GPL FUD is that GPL code cannot be combined with code under different licenses, which is wrong.
The fact that SUN uses GRUB for both OpenSolaris and ClosedSolaris proves that SUN’s anti-GPL FUD is wrong, GPL code can easily combined with code unter different licenses and now it’s finally proven.
SUN should use GRUB and in return stop its anti-GPL FUD, this would be the best solution for all parties.
Ah the standard Linux answer to “big iron” is clusters, not every application is cluster aware. Most of the applications (other than Oracle) that we use are not cluster aware. So the “cheap x86 cluster” argument doesn’t fly. And that “performacne” you talk about is also on specialized workloads, again not the every day stuff that big iron is used for.
It’s not just about saving money, there is the reliability issue. I don’t expect most PC servers to run as long or as reliably as a SPARC. And considering that these servers will probably be in place for 7 to 10 years, the places who buy this kind of hardware as a capitol investment.
If I wanted the ultimate in fault tolerance, I would have clustered SunFire machines using either Sun or Veritas Cluster software, I could (not recommended by Sun) split my 4800’s into 2 domains and using two machines, cluster the domains.
And what about IBM and HP, they sell similar hardware. So using your argument, customers who use IBM pSeries, zSeries and HP’s SuperDome and Integrity servers should replace them with PC’s running Linux? This hardware is just as expensive (if not more so) than Sun.
To provide the services we require (several large Oracle databases, document management, and a development environment) would probably run on PC servers, poorly. yes, it can be done with PC servers running Linux or Windows (Microsoft uses the same answer to address the shortcomings of PC hardware). Or I can use Solaris, AIX, or HP-UX on Enterprise grade hardware. Does it cost more, yes it does, is it more reliable, YES it is. And that is what people who spend the “big bucks” want.
…that Sun would choose for a system running on enterprise-class machines a bootloader whose RAID support is problematic, to say the least.
Hah, the cluster answer. It’s always obvious that the person who gives the cluster answer has never worked on a cluster.
Yeah and I think I know what goes with cluster and it begins with F!
Read the subject line, nowhere did it say SPARC. And I am sure that Sun is addressing the issues with grub and SVM (if there are any).
I’ve worked with several clusters, the first one I setup was around 1998. One cluster option is to use a Localdirector or even a trick in DNS to load balance across a group of PCs. Another is to use software like Oracle or Trolltech’s Teambuilder or OpenMOSIX to provide load balancing, fault tolerance and redunancy, etc.
But if your decision to use “big iron” is based on your personal preference or trust in Sun hardware or some boost to your ego for spending the “big bucks” nothing I say will change your mind.
The guy didn’t ask about Linux on a Sunfire 4800.
And, clusters do not offer a satisfactory solution for a very large class of softwares. Clusters are way overhyped, due in large part to PHBs not understanding Amdahl’s law, multithreading and the programmatic problem at hand. They just hear “many CPUs” and they are happy. Sadly Linux caters to such morons.
No, its about using the “right tool for the job”, and I don’t see that clustering a bunch of PC’s is better than big iron. Because if that was really true Sun, IBM, HP, SGI, NEC, Cray, and a host of others would be in big trouble. And since IBM and HP are porting Linux to their big iron hardware, it seems they are trying to “guild the lily” by providing Linux for their high-end stuff in order to give their potential customers a “choice”.
Also your comment about using obsolete hardware tells me you are willing to sacrifice performance since older hardware is not likely to perform as well as the newer stuff. I spent enough time when I was in the Navy using the “spit and bailing wire” method of keeping equipment up. I don’t have to anymore, and neither should a customer.
When reliability and proven (not theoretical) performance is more important, some customers prefer to spend the money for quality equipment and services. I work for customers who want to save money, but know what hardware and software they need to get the performance they want. And the savings comes as a result of getting the job done with minimum downtime. Not by buying cheap hardware and expecting it run like the “expensive stuff”.
HP, SGI, and Sun’s financial trends should certainly worry someone.
No, its about using the “right tool for the job”, and I don’t see that clustering a bunch of PC’s is better than big iron. Because if that was really true Sun, IBM, HP, SGI, NEC, Cray, and a host of others would be in big trouble. And since IBM and HP are porting Linux to their big iron hardware, it seems they are trying to “guild the lily” by providing Linux for their high-end stuff in order to give their potential customers a “choice”.
Well, Sun’s financials aren’t exactly something to brag about mostly because they’re having a hard time selling the high margin big iron. Why do you think they now sell glorified PCs (aka their Opteron line)? SGI is damn near out of business and what’s left of them sells high end clustered super computers (which they like to sell with Linux installed on). Cray is big on clustered supercomputers as well these days. Given that you are trying to say that big iron is different from clustered supercomputers, I wouldn’t use SGI or Cray as an example. IBM is more of a services company then a hardware company but both HP and IBM will preinstall Linux on their big iron systems. Do you have any examples of companies that actually are doing well selling big iron systems that won’t install Linux on them for you?
Also your comment about using obsolete hardware tells me you are willing to sacrifice performance since older hardware is not likely to perform as well as the newer stuff. I spent enough time when I was in the Navy using the “spit and bailing wire” method of keeping equipment up. I don’t have to anymore, and neither should a customer.
Depending on the job I’d happily take obsolete hardware over expensive “big iron”. Who needs $15,000+ Sun systems for web heads when a $500 Dell throw away 1U does the job just fine? Even into the midrange, why pay big bucks for a Sun SPARC system when an IBM xSeries quad Xeon or whatever they call their new Opteron systems are faster and cheaper and have similar quality to the equivalent Sun systems?
When reliability and proven (not theoretical) performance is more important, some customers prefer to spend the money for quality equipment and services. I work for customers who want to save money, but know what hardware and software they need to get the performance they want. And the savings comes as a result of getting the job done with minimum downtime. Not by buying cheap hardware and expecting it run like the “expensive stuff”.
If someone is looking for reliability *and* performance, SPARC is the last place I’d send them. IA64 and Power both smoke the best that Sun or Fujitsu has available today. What is really sad is that an HP SuperDome running Win2k3 will handily outperform the best that Sun has to offer when it comes to stuff like databases. And if you need to absolute highest levels of I/O and reliability, you can’t beat a zSeries.
There really is a reason that Sun is having a hard time making a profit on their hardware, it just isn’t that good anymore.
The Integrity is HP’s way of trying to salvage their messed up strategy at the Enterprise level. Ask someone who is an existing HP customer if they want to get rid of their PA-RISC system to get Itanic, because that is what HP wants them to do. Wasn’t someone at HP not too long ago bitching about Sun “takin gaway their customers”? Yeah, people are lining up to buy the Integrity servers to run Windows, even you said it “when a $500 Dell throw away 1U does the job just fine”, the same people won’t spend that kind of money on an Integrity system because it’s not cheap!
One of the guys I work with at his previous place bought IBM xSeries, his view is they are junk! Constant replacing of hardware and minimal uptime. I have experience with IBM as well (AIX pSeries), it’s nice gear but not that nice. And have you priced those POWER 5 systems lately?
Yeah, some idiot bought us one of those Dell 2650’s, with no spare disks, no redundant power. Oh yeah, that’s what I want in my data center, NOT! Cheap is cheap, and I have heard this far too many times from PHB’s “ooh, look at this, we can save lots of money”. And the poor sysadmin has to figure out how to make it work.
As always the “financials” seem to come up, as if that is really a measure of system performance. Why don’t you save that for the “BSD is dying” troll?
Is the performance of the zSeries great because its a mainframe, or because you can run Linux in an LPAR? Unfortunately everything is NOT about Linux.
If you choose to buy whatever you want do so! For 12 years I have used Sun and intend to do so in the future. Just as the Linux crowd is loyal, so am I.
In response to the following post –
Somewhat odd that Sun would choose for a system running on enterprise-class machines a bootloader whose RAID support is problematic, to say the least.
– Robert Escue wrote:
Read the subject line, nowhere did it say SPARC. And I am sure that Sun is addressing the issues with grub and SVM (if there are any).
Umm, Robert, nowhere did *I* mention SPARC or SVM. I did mention “enterprise-class machines,” but there are certainly non-SPARC enterprise-class machines. Nor was I referring to SVM when mentioning RAID. I meant to include garden-variety hardware RAID, and I certainly meant to include machines with x86 processors.
Now instead of criticizing something I didn’t say, why don’t you try to install GRUB on any RAID system of your choosing, and report back to us on how entertaining the experience was.
Well considering that the V20z and V40z isn’t what I would call an “Enterprise Class” machines. Maybe what you need to do is be more specific, because we are using RedHat Enterprise Linux on our V20z’s (which use grub) and the onboard LSI RAID doesn’t seem to have any problems.
Is this what you’re talking about?
http://support.novell.com/techcenter/tips/10062.html
http://support.novell.com/cgi-bin/search/searchtid.cgi?/en/2004/04/…
http://64.233.187.104/search?q=cache:3gGG2-G-Qi0J:www.webservertalk…
I don’t exactly call IDE RAID “Enterprise Class” stuff, I prefer SCSI and Fibre Channel myself.
> If someone is looking for reliability *and* performance, SPARC is the last place I’d send them. IA64 and Power both smoke the best that Sun or Fujitsu has available today
There is a lot more to the servers than just CPU performance, plus it depends on how you define performance. If you shop for servers based on the processor benchmarks, you’re a fool, plain and simple. There are a lot more to be taken into consideration than some stupic SPECint/SPECfp benchmark. Yes, based on SPEC benchmarks SPARC is behind Itanic and Power, but at the same time SPARC is hell of a lot better value for the money in terms of investement protection and overall 64-bit story. Let me explain why.
With Itanic it is plain and obvious, the lack of binary compatibility, lack of good compilers and brain share already brought Itanic to canatonic state and I won’t be surprized if Itanic completely disappears in just a few years. Even HP lost its faith in IA64 and is now happily sells Opteron. IA64 SuperDome can’t really deliver and if you didn’t know 90% of all SuperDomes sold are still PA-RISC. I wonder what is going to happen to HP-UX when PA-RISC is completely EOL’ed (hint – dead).
With Power5 the story is not as simple but Power is still far from being perfect as lots of bigots claim it be.
* First off, raw performance wise the processor is pretty good, but even there IBM couldn’t play it straight. The SPEC benchmarks you see for Power are about 10 to 20 percent bullshit — the highest Power SPEC benchmarks are done on MCM’s (multi-core modules) and IBM used a dirty trick by turning off three out of four cores on the MCM giving all of the cache on the MCM to just one core, thus artificially improving the results. You can safely reduce the claimed SPEC results by about 10 to 20 percent to get the real world performance.
* Second, it looks like IBM didn’t get the memo about how much binary compatibility actually matters in this world — AIX on Power have been breaking binary compatibility with each generation of Power processor forcing major recompiles and box swaps. It looks like by IBM’s book the forklift upgrades are a day-to-day reality it seems (LOL). There is nothing like that with Sun — you can mix different family CPU’s in the same frame without any problems really extending the life your server and saving you $$$. Plus every break in binary compatibility reduces the number of applications available on the platform. Did you know that there are less than 200 certified application for Power5? You need to recompile to take advantage of Power5. Compare that with a few thousand of applications certified for SPARC (SPARC never broken binary compatibility in history).
* Third, micropartitioning on Power5 seriously sucks ass as it will sap almost 50% of CPU in just hypervisor overhead if you’ve got more that 10 micropartitions on the processor. With Solaris 10 zones there is no overhead whatsoever and a potential to run literally hundreds of virtual servers per CPU.
* Forth, AIX is grotesquely overpriced piece of garbage that is already so hopelessly behind Solaris that it is not even funny. Solaris is an absolute steal compared with AIX. IBM charges you about $1700 per CPU on averarge (depending on the box) and will charge you more for every LPAR or micropartition on top of that. Just to give you an example, AIX licensing on a 32 processor box comes just over $80,000 the last time I checked. Compare that with Solaris that can be had for free!
* Fifth, AIX on Power is still behind Solaris on SPARC in terms of RAS and DR (Dinamic Reconfiguration) in particular. Under AIX you still have to shutdown/reboot the entire box to perform maintenance on anything related to the CPU/memory, on Solaris/SPARC on the other hand you can replace uniboards on the live system without any problem whatsoever.
> why pay big bucks for a Sun SPARC system when an IBM xSeries quad Xeon or whatever they call their new Opteron systems are faster and cheaper and have similar quality to the equivalent Sun systems?
Err, UltraSparc powered Sun v440 with 4 processors is about the same price as 4 processor xSeries IBM crap. Only with Sun you buy the best proven and tested 64-bit platform, whereas with IBM you buy a glorified PC. Do you see a difference? Sun UltraSparc boxes are very competitive nowadays, regardless of whatever PC-minded morons tell you.
I’m sure the Linux crowd doesn’t want to hear what you just said, since they thrive on benchmarks regardless of whether they are based in reality or not. The POWER 4 based systems I worked on in a previous job weren’t all that fast, but they sure were expensive. I figured it would be pretty much the same with POWER 5.
HP is stoned if they think their high-end customers are going to pay to port their apps from PA-RISC to Itanium. And despite the fact that Linux runs on a SuperDome, I really can’t see a company spending a million dollars on a basically unproven platform to run Linux.
And like I have said before, if IBM and HP were “really” into helping Linux, they would “give to the community” the ability to create LPARs and vPars respectively to Linux. Don’t bank on that happening any time in the future. What’s the point in running on big iron if you can’t utilize big iron features.
Amen!
> If someone is looking for reliability *and* performance, SPARC is the last place I’d send them. IA64 and Power both smoke the best that Sun or Fujitsu has available today
There is a lot more to the servers than just CPU performance, plus it depends on how you define performance. If you shop for servers based on the processor benchmarks, you’re a fool, plain and simple. There are a lot more to be taken into consideration than some stupic SPECint/SPECfp benchmark. Yes, based on SPEC benchmarks SPARC is behind Itanic and Power, but at the same time SPARC is hell of a lot better value for the money in terms of investement protection and overall 64-bit story. Let me explain why.
I never said there wasn’t more to servers than CPU performance, I said that if you want reliability and performance you don’t go with SPARC. None of your excuses changes that either, SPARC is slow. Perhaps you should spend less time defending a chip which even Sun can’t sell (see their Opteron boxes for more details) and start asking how they are going to save that sinking ship.
The Integrity is HP’s way of trying to salvage their messed up strategy at the Enterprise level. Ask someone who is an existing HP customer if they want to get rid of their PA-RISC system to get Itanic, because that is what HP wants them to do. Wasn’t someone at HP not too long ago bitching about Sun “takin gaway their customers”? Yeah, people are lining up to buy the Integrity servers to run Windows, even you said it “when a $500 Dell throw away 1U does the job just fine”, the same people won’t spend that kind of money on an Integrity system because it’s not cheap!
There are about the same number of people lined up to buy Sun’s big iron as HP’s (or IBM’s for that matter). Doesn’t change the fact that a SuperDome running Win2k3 would provide better performance for, for example, a database system then the comparable Sun system. Sun is getting so far behind in the performance game that it’s silly to even look at their gear if you aren’t already an all Sun shop.
One of the guys I work with at his previous place bought IBM xSeries, his view is they are junk! Constant replacing of hardware and minimal uptime. I have experience with IBM as well (AIX pSeries), it’s nice gear but not that nice. And have you priced those POWER 5 systems lately?
And I work on 2 xSeries and I’ve had no problems with either of them. 24×7 solid for over 2 years now. As for the price of a pSeries, you’re paying big bucks for some of the highest quality hardware you can buy. Isn’t that your whole argument, you pay more for better gear?
Yeah, some idiot bought us one of those Dell 2650’s, with no spare disks, no redundant power. Oh yeah, that’s what I want in my data center, NOT! Cheap is cheap, and I have heard this far too many times from PHB’s “ooh, look at this, we can save lots of money”. And the poor sysadmin has to figure out how to make it work.
Again, depends on what you need.
As always the “financials” seem to come up, as if that is really a measure of system performance. Why don’t you save that for the “BSD is dying” troll?
I didn’t bring financials up, you did when you said “Because if that was really true Sun, IBM, HP, SGI, NEC, Cray, and a host of others would be in big trouble.” Guess what, Sun, SGI and Cray are all in varying amounts of trouble because people aren’t buying the big iron. HP, IBM and NEC aren’t because they’re so diversified.
Is the performance of the zSeries great because its a mainframe, or because you can run Linux in an LPAR? Unfortunately everything is NOT about Linux.
Depends on what you mean by performance. They aren’t the fastest number crunchers out there. I said if you want I/O and reliability you buy a zSeries. (As for your LPAR comment, everything except z/OS and a handful of native apps run in an LPAR, that’s just how a zSeries works).
why pay big bucks for a Sun SPARC system when an IBM xSeries quad Xeon or whatever they call their new Opteron systems are faster and cheaper and have similar quality to the equivalent Sun systems?
Err, UltraSparc powered Sun v440 with 4 processors is about the same price as 4 processor xSeries IBM crap. Only with Sun you buy the best proven and tested 64-bit platform, whereas with IBM you buy a glorified PC. Do you see a difference? Sun UltraSparc boxes are very competitive nowadays, regardless of whatever PC-minded morons tell you.
Sun boxes might be priced competitively but you surely jest if you think they’re actually competetive in any other way. (Best proven and tested 64 bit platform? Who are you kidding. I just hope Sun chip designers stay away from the Opteron, it’d be sad if they manage to ruin two architectures)
And the IBM xSeries is hardly a glorified PC, if you’d ever used one you’d know that IBM uses all their own gear to make an enterprise level system using Intel processors. This in comparison to Sun’s Opteron line which are glorified PCs. They’re like Dells, basic whitebox gear in a pretty Sun case.
Sorry I’m offending all the Sun/SPARC lovers out there, but don’t shoot the messenger. Go ask Sun why their hardware performs so dismally. Go ask Sun why, if they think SPARC is so great, that they’ve had to release a line of Opteron systems (which, btw, is going to kill off the last vestigages of their low and midrange SPARC systems).
> I never said there wasn’t more to servers than CPU performance, I said that if you want reliability and performance you don’t go with SPARC. None of your excuses changes that either, SPARC is slow.
Fist of all, reliability should be off your list entirely. As I pointed out above SPARC systems from both Sun and Fujitsu are actually quite a bit ahead of both IBM and HP in terms of RAS. In respect to performance, yes, UltraSparc currently has some issues in respect to per-core single thread performance, but Sun can already equalize the per-core disparity by stuffing more cores on die and stuffing more chips into the system than their competitors. Sun FirePlane based systems scale more linearly than either IBM or HP and have no data skew, so having many processors on the backplane have less of decreasing marginal returns effect on overall performance. If single core performance is really your thing, than ask for Fujitsu PrimePower with SPARC64, the 2Ghz chip has single thread performance pretty much on par with the latest Power5 offering. Plus with Sun and Fujistu joining their chip development operations into APL by the end of this year, the “SPARC is slow” garbage is going to become history anyway.
On the side note, besides everything said above, as they say “you date hardware, but you marry your OS”. You can flirt with hardware all you want, but the real choice should come down to the operating system. And on that front Solaris 10 is the OS you would want to marry, not AIX or HP or Linux or god forbid Windows. Solaris 10 is head and sholders above the rest.
> There are about the same number of people lined up to buy Sun’s big iron as HP’s (or IBM’s for that matter). Doesn’t change the fact that a SuperDome running Win2k3 would provide better performance for, for example, a database system then the comparable Sun system. Sun is getting so far behind in the performance game that it’s silly to even look at their gear if you aren’t already an all Sun shop.
I’m not sure where you get these ideas that Sun somehow hopelessly behind HP and IBM in databases, I hope you didn’t drink too much of that IBM koolaid with whatever they put in it. Just as sort of a reality check consider this, 2 out 3 Oracle deployments still go on Solaris/Sparc. The biggest data warehouses in the world run on Solaris/Sparc. And SuperDome with Win2k3, are you kidding? This is just laughable! I bet HP haven’t sold a single one in that configuration. As far as database applications are concerned (especially highly transactional) it is not about CPU performance, it more about memory bandwidth and memory access symmetry. If you need more convincing why Sun systems are still the best for database applications and data wharehousing in general, may be this whitepaper from the creators of the biggest data wharehouses in the world can enlighten you:
http://www.clickstreamconsulting.com/DW_Architectures.pdf
> Go ask Sun why their hardware performs so dismally. Go ask Sun why, if they think SPARC is so great, that they’ve had to release a line of Opteron systems (which, btw, is going to kill off the last vestigages of their low and midrange SPARC systems).
If unit sales are of any indicator, SPARC systems are still very popular and aparently many customers still love them (yours truly included). According to the report from Sun, the unit shipments for SPARC gear increased by 20 odd percent in the last year alone. It doesn’t look it is whithering away, that’s for sure. And why do you thing the fact that Sun started selling Opteron has any inclination to SPARC? Sun is trying to grab as much as possible at the low end to improve market penetration. Plus, since Sun is not just a box dropper any more and sells software, storage and services, if they can attach a middleware or storage sale to some low end Opteron kit even if it ends up running Linux, Sun still wins. They try to diversify from just being a narrowly focused server vendor and this is a good thing.
Have you actually used a V20z or a V40z, or are you just talking trash? Just two feet away from me is three V20z’s (of the ten we purchased) and I hardly consider them PC’s. There are some things I don’t like about them, but I would rather have them than Dell boxes.
And just who is going to spend $250,000 (2003 dollars) on a minimally configured Intergrity SuperDome to run Windows?
No when it comes to IBM gear, it is just high priced. And I just love getting a “crazy 8’s” over how you plug a cord into a power supply!
It’s pretty simple. If they cannot sell their hardware then they will cease to exist. I don’t think you’re particularly qualified to discuss performance issues (at least from every single post I’ve read of yours, your ability to discuss technical performance issues seems to treat your conclusions as axiomatic and thus not particularly interesting), so I don’t really feel like being part of your conversation.
If there is such a tremendous demand for their hardware and what you think it does, then I suggest that you shore up your suppliers’ financials for them by purchasing it. You seem to think that these companies shouldn’t be worried, when their financial trends indicate that if they aren’t they’re not going to be around, or they’re going to have to cut expenses to reflect selling niche products. That’s basically what SGI has been doing. Who knows what HP and Sun are going to decide on doing; I doubt they do.
Well if you don’t like my discussions on performance, don’t particpate! I don’t mind (as I am sure others here won’t as well). Maybe you should read my posts at http://www.sysadmintalk.com, where system administrators help each other. Unlike the mostly FUD filled discussions here, at sysadmintalk we actually talk about REAL issues, like Oracle with and without DIRECTIO on disks (amongst other things). Unfortunately real performance topics are not discussed here.
If you want to talk about specific performance issues, fine. I think I’m more than capable, are you?
In his second and third responses to my following post –
Somewhat odd that Sun would choose for a system running on enterprise-class machines a bootloader whose RAID support is problematic, to say the least.
– Robert Escue said:
Well considering that the V20z and V40z isn’t what I would call an “Enterprise Class” machines. Maybe what you need to do is be more specific, because we are using RedHat Enterprise Linux on our V20z’s (which use grub) and the onboard LSI RAID doesn’t seem to have any problems.
– and:
Is this what you’re talking about?
http://support.novell.com/techcenter/tips/10062.html
http://support.novell.com/cgi-bin/search/searchtid.cgi?/en/2004/04/…..
http://64.233.187.104/search?q=cache:3gGG2-G-Qi0J:www.webservertalk…..
I don’t exactly call IDE RAID “Enterprise Class” stuff, I prefer SCSI and Fibre Channel myself.
Your first response is to say that two machines I never mentioned aren’t “enterprise class.” Umm, fine, except it seems entirely irrelevant to the subject at hand. You also state that GRUB appears to run fine on some machines where you work that have LSI RAID. Did you set up GRUB on these machines? If so, please tell us the steps you took, and if you would characterize the process as easier or more difficult than average.
Your second response cites one instance of difficulty installing GRUB on IDE RAID, and two notes saying *the 2.6 Linux kernel* will no longer support certain software RAID cards. You then extrapolate from these to a remark that you don’t consider IDE “enterprise class.”
In comments responding to others in this thread you’ve praised your own technical competence. Thus I find it puzzling that you’ve apparently confused RAID support in the Linux kernel with problems in the GRUB software application related to RAID. It’s also disconcerting that you would, from a single notice of GRUB problems with IDE RAID, jump to the conclusion that this problem is restricted to IDE. It’s not. You can easily confirm this with some quick searches relating to GRUB, RAID, and SCSI.
Well, firstly let me say that I find sysadmintalk.com an excellent, very pertinent resource. Also, I have been working with HP, IBM and Sun hardware (and HP-UX, Linux and Solaris) for almost a decade now. Also, I have no reason to doubt Robert’s technical expertise, quite the contrary, I think he’s right on the money in what he says here.
However, said all that, I am not a big fan of GRUB. There are several things I dislike about it, and really didn’t feel confortable working with it.
Well, I am sure Sun has real expert making these kinds of decision – I am nowhere qualified enough to make judgement of any kind, about x86 Solaris bootloader.
Yes grub was set up on all of the V20z’s with no issues at all. Just install as you would on a machine without RAID. And our machines have survived multiple reboots with no loss of drives.
Now you want specifc answers, post specific questions instead of your blathering bullshit.
Thanks for the props!
Hi,
What specific performance issues do you have with SPARC (Sun
and Fujitsu)? Are you just using spec scores to make your
judgements or is there something else?
If you are talking about single thread performance on mostly
cache resident data I would have to agree, at the moment Sun
SPARC is not competitive (Fujitsu SPARC is however). IBM
would absolutely love you to make your hardware decisions
on this premise.
However, real customers with real workloads usually find
that performance is a lot closer than what set benchmarks
and benchmarketing would suggest.
The best thing to do is call Sun and get YOUR app with
YOUR data running at one of the customer benchmark centres.
If you are having power/cooling/real estate problems then
SPARC will be your answer with Niagara. Don’t write it off
just yet.
Hey, what happened with Niagra!? Trying to be politically correct, are we ;oP
Did you rewert to NiagAra after marketing found out? ;o)
“it is additional evidence that the only relationship they want with the free software community is one where they take and don’t give back”
What about Open Office? NFS? RPC? Their contributions to Gnome? What do you think Open Solaris is all about? (and what about all the patents they donated?) Snu has made major contributions to OSS and these continuing claims that they haven’t are pure FUD.
I have no idea what they want, but obviously the contributions Sun has made is not enough. I am also trying to figure out the animosity that the Linux zealots have about Sun and Solaris. Most of the arguments I have read here are absolutely ridiculous like “I don’t like what Jonathan Schwartz has to say”. Like that really affects my decision making process as to whether or not to use a product.
I would just like to see an article or link that has anything to do with Solaris not become a FUD fest. You would think the Linux crowd would be more focused on trying to persuade others to the benefits of Linux online and off, rather than anger those they might have to spend more time convincing. Or at least that’s what I did when I was part of “Team OS/2”.
I think you missed the biggest issue; the latency issue on large POWER boxes when compared to SPARC. Now, sure, I’m all for bashing *UltraSPARC*, thats an easy target, but price/performance wise, SPARC64 isn’t half bad. IMHO SUN should replace the UltraSPARC III cores in their UltraSPARC IV with SPARC64, and release it as a UltraSPARC V, that alone, dual core, running at 2.1Ghz per core would provide a damn good power boost to their enterprise line up.
Oh, and just as a side issue, SPARC is an openstandard. Unlike POWER which is controlled by IBM – oh, and don’t believe the OpenPOWER hype – they’ve only released a *SMALL* subset of the ISA to the market, and even then, there are BIG strings attached to it.
“I think you missed the biggest issue; the latency issue on large POWER boxes when compared to SPARC. Now, sure, I’m all for bashing *UltraSPARC*, thats an easy target, but price/performance wise, SPARC64 isn’t half bad. IMHO SUN should replace the UltraSPARC III cores in their UltraSPARC IV with SPARC64, and release it as a UltraSPARC V, that alone, dual core, running at 2.1Ghz per core would provide a damn good power boost to their enterprise line up.”
It’s called the APL. POR is 2006.
In response to my observation that I found it odd Sun would adopt GRUB as its bootloader for enterprise-class machines in view of some problems GRUB had with RAID, Robert Escue’s latest reply was:
Yes grub was set up on all of the V20z’s with no issues at all. Just install as you would on a machine without RAID. And our machines have survived multiple reboots with no loss of drives.
Now you want specific answers, post specific questions instead of your blathering bull***t.
I don’t recall blathering, nor asking questions, simply making an observation. Regarding specifics, GRUB has had and continues AFAIK to have problems identifying, accessing, and being installed to IDE or SCSI disks in RAID arrays (other than RAID1, where installation at least in certain Linux distros is now rather routine).
A decent non-technical explanation appears in this March 2005 e-mail to the bug-grub mailing list:
http://lists.gnu.org/archive/html/bug-grub/2005-03/msg00059.html
I am also trying to figure out the animosity that the Linux zealots have about Sun and Solaris. Most of the arguments I have read here are absolutely ridiculous like [etc…]
Of course, if weren’t for Linux zealots, all would be peace and harmony, wouldn’t it? NOT! Maybe if there were no zealots, perhaps. Every advocacy group has zealots and they are usually rude, stupid, fatuous and self-serving. Sun does not help the matter with its own brand of FUD.
I would just like to see an article or link that has anything to do with Solaris not become a FUD fest.
That would be remarkable, wouldn’t it?
You would think the Linux crowd would be more focused on trying to persuade others to the benefits of Linux online and off…
And you would think that “the Solaris crowd” would be more focused on trying to persuade others of the benefits of Solaris online and off rather than telling me how bad Linux is, how inferior it is and, oh, by the way, how naughty Red Hat is. Solaris and its advocates would be better served if they would busy themselves with showing why Solaris is good and to leave off trying to tell me to hate Linux. Failing that, they can at least shut up, along with all other zealots, of course.
… rather than anger those they might have to spend more time convincing. Or at least that’s what I did when I was part of “Team OS/2”.
Maybe you did. However, I, too, was a part of Team OS/2, and was well acquainted with the good, the bad and ugly of that team effort. I would like to think that the nebulous group which umbrella’d under the appellation “Team OS/2” was better than most advocacy groups, but it wasn’t, not on average. As usual, only individuals stood out as measured and reasonable, whereas others were merely loud and vituperative, just like every other advocacy group.
We are often less critical of those whose views are more congenial with our own and much more so of those who disagree. After all, since we’re right and they’re wrong, we are only speaking in the tones of righteousness whereas they are being shrill and abusive. It was ever thus.
Harking back to the days of Windows vs. OS/2 is both amusing and pathetic since you (presumably) liken Linux to Windows (the bad guys) and Solaris to OS/2 (the good guys), when not so long ago, the face-off was reckoned to be the Windows bad guys vs. the Linux good guys.