The Chicago Mercantile Exchange has been migrating its servers from Solaris to Linux over a period of time with about 30% completed now “When the Merc began considering its Linux deployment more than a year ago, Sun knew it wasn’t in a strong position to compete. ‘We didn’t have a good answer for them,’ acknowledges Glenn Weinberg, vice president of the operating platforms group at Sun.”
“The CME’s architecture consists of three major technologies: Unix and Linux systems handle the input paths for orders and output paths for quotes. Hewlett-Packard NonStop servers send out order confirmations and quotes. And IBM mainframes are responsible for “clearing,” or processing, of all trades.”
Basically they are just using Linux/UNIX for their webservers. Anything serious, they go straight for the big iron.
“But he says he thinks the Merc’s Linux vendor, Red Hat, needs to improve its support. Currently, he says, Red Hat emphasizes purchasing more products as a way to fix problems. “When there are issues, they need to step up better,” Panfil says.””
RedHat is still a pretty small company trying to compete against IBM and Sun full-service support teams.
“But he says he thinks the Merc’s Linux vendor, Red Hat, needs to improve its support. Currently, he says, Red Hat emphasizes purchasing more products as a way to fix problems. “When there are issues, they need to step up better,” Panfil says.”
I couldn’t agree more, RedHat’s support absolutely stinks, which is not surprising as RedHat doesn’t actually own any technology they try to peddle to the customers — one of the drawbacks of trying to sell the open-source tech you didn’t develop in house. I logged a few support calls with RedHat and each time I would get an impression that the support people at RedHat don’t know much about the product they are trying to support — they always try to give some stupid reference to some public project web site or a forum, which raises a huge question — what are they charging all that support money for? I seriously doubt RedHat will ever be able to adequately support the OS. If you want good support, stay with Unix (AIX, HP-UX, Solaris).
> The Chicago Mercantile Exchange has been migrating its servers from Solaris to Linux over a period of time with about 30% completed now…
I’m sure it is going to be a very regretful move considering all new features and opportunities they are going to miss in Solaris 10 — Solaris 10 beats RedHat Linux on all counts from price and features to raw performance. I have to say it is a little stupid to forgo all the beautiful technology in Solaris 10 and just going with the Linux hype.
I’m sure it is going to be a very regretful move considering all new features and opportunities they are going to miss in Solaris 10 — Solaris 10 beats RedHat Linux on all counts from price and features to raw performance. I have to say it is a little stupid to forgo all the beautiful technology in Solaris 10 and just going with the Linux hype.
What’s that you say? Solaris 10 not overhyped? You must be a Sun fanatic.
I’d be very interested as to where Solaris 10 beats Linux in “raw performance” though. Links, please (no crap from sun.com, of course).
This article could have been titled. Solaris vs. Linux: Great flame ensues.
Can we agree on the following? Both Solaris and Linux are operating systems with a great conceptual pedigree: Unix.
Linux is newer, yet has gained the momentum it has through technical merit, performance and by being able to deal with increasingly more complex jobs.
If somebody thinks that Amazon’s purchasing, tracking and billing system isn’t complex, then you are beyond silly. Could they have used Solaris or another OS? Quite likely.
Would they have gotten the same price/performance ratio? Definitely not.
> I’d be very interested as to where Solaris 10 beats Linux in “raw performance” though. Links, please (no crap from sun.com, of course).
Check it for yourself, download Solaris 10 and test your applications against Linux. I tried it myself and it looks like Sun’s claims are pretty much on the money. The official benchmarks will most likely be released with the official availability of Solaris 10 at the end of this month.
> Solaris 10 not overhyped?
Solaris 10 is not overhyped. The new features in Solaris 10 that get the spotlight are quite revolutionary and don’t have any equivalent in any other server OS including Linux (dtrace, ZFS, zones, etc.).
> Could they have used Solaris or another OS? Quite likely. Would they have gotten the same price/performance ratio? Definitely not.
You can get better price/performance from Solaris running on the same x86 hardware and give you lower TCO — Solaris has much more superior resource management facilities allowing you to consolidate and sqeeze out more out of your hardware. Solaris 10 with zones will put Linux to shame on that metric — you will be able to cheaply consolidate hundreds if not thousands of servers onto one machine with zones ($$$ saved).
Solaris 10 is not overhyped. The new features in Solaris 10 that get the spotlight are quite revolutionary and don’t have any equivalent in any other server OS including Linux (dtrace, ZFS, zones, etc.).
Well, to be fair, Linux does have its own version of DTrace called “KProbes” which was implemented by HP and made it into 2.6.9. There are other virtualization mechanisms which accomplish the same ends as Zones. However there isn’t anything quite as powerful or featureful as zones… there’s UML which is monumentally slow and Xen which doesn’t support multiple processors, AMD64, or SPARC. Zones use seperate kernel instances which communicate via a high performance kernel mode interface, and can be bound to certain system resources in a manner not possible on Linux (You can designate a certain number of processors/processor shares to bind a Zone to using the Fair Share Scheduler)
What you can’t argue with is that Solaris is substantially more optimized and has a better design than Linux. Solaris implements an mbuf model which was copied by FreeBSD with UMA, where Linux’s VMM interface has fundamental design errors which necessitate hacks like the OOM killer. While Linux has kernel preemption, it substantially harms kernel throughput, whereas Solaris is able to achieve both the low latency of kernel preemption and high throughput by demarking non-preemption zones within the kernel where kernel throughput would be harmed.
But don’t take my word for it. Soon the forum will be beseiged by a hundred Linux zealots, who believe that the bazaar development model can somehow produce a better high level design than professionally written software, and that given the sheer number of programmers involved with Linux it must be inherently better! (That’s what Fred Brooks said right? More programmers are better right? In his book the Factual Man-Month… oh wait, or was that the Mythical Man-Month…)
I contend in a large project like an OS kernel less is more… a cohesive experts group which has been with the code since its inception will, in general, produce better code than having random people from time to time cobble their own modifications onto the existing codebase, originating new features through trial and error (i.e. the round-robin I/O scheduler which has been exhibiting… unusual performance characteristcs) rather than through a professional research.
Of course point out that Linux development has been mismanaged (Rik van Riel certainly contends that, http://linuxtoday.com/developer/2002011501620INKN) and you’ll get the typical OSS idealism which they think they can hold up to the practicality of the business world. Lacking real world experience with Solaris, they’ll link to a few poorly conducted, unscientific benchmarks. But don’t try to link to any benchmark Sun endorses, clearly it’s biased because Linux didn’t win.
I don’t know where this attitude originates from. It certainly annoys me greatly. Solaris has been the performance leader among Unix operating systems for quite some time (especially for database usage), and until they were forced to deal with HP hawking HP-UX, Tru64, and Linux (if you want to talk about a company with a confused business strategy…) they were the leader in Unix server sales. With Fujitsu manufacturing SPARC64 processors in excess of 2GHz, SPARC certainly has a bright future. I’m not really sure why it was decided to release the Solaris source or if it will really have any impact on Solaris development at all, but it can no longer be said that if you want to know how Solaris implements a particular set of functionality, you can’t see because Solaris is a black box.
Solaris 10 is a revolutionary release, and ensures that Solaris will continue to be the performance leader for the near future, but rest assured that a zealously touted, mismanaged, cobbled together open source operating system which continues to get basic Solaris features 5 years after they were implemented in Solaris will continued to be touted by its brainwashed lovers as the performance leader simply because there is no way within their ideology that Linux could not have surpassed Solaris by this point… and no amount of benchmarks, accounts of real world experience, or code analysis will ever sway them from this viewpoint.
Linux is newer, yet has gained the momentum it has through technical merit […]
Say what ? Which “technical merit” are you thinking of ? Linux has only gained parity in features in the last 12 – 24 months most of the alternatives have had for the better part of a decade, if not longer (to say nothing of waiting for it to mature).
Linux has gained momentum largely through grassroots overhyping and zealotry. You’d have a pretty hard time finding any contemporaries it is (or ever was) “technically” superior to.
> Well, to be fair, Linux does have its own version of DTrace called “KProbes” which was implemented by HP and made it into 2.6.9.
It has been discussed on these forums for a number of times and DTrace engineers talked about this — DTrace is fundamentally different and infinitely more flexible than KProbes. Read DTrace documentation for more information. So far there is nothing out there regardless of the OS that can match DTrace.
It has been discussed on these forums for a number of times and DTrace engineers talked about this — DTrace is fundamentally different and infinitely more flexible than KProbes.
That was the whole point of my post… which apparently you missed. Linux advocates scrape for feature parity comparisons neglecting that Sun’s implementations are substantially better… the perform better and give the administrator more flexibility.
And yes, KProbes are static entry points which reduce performance even when not in use whereas DTrace is a dynamic facility which has no impact on performance when it’s not in use…
Zones don’t require a virtual machine, unlike Xen, and use a high performance kernel mode interface between kernel instances, rather than UML which, obviously, runs in user mode and requires context switches to talk to the parent kernel.
And so on and so on…
Bravo!
Great post!
However I can assure you the “itchy-people” won’t tolerate this kind of honestly for long.
Also can you like to some more information on ZFS, and Zones please?
Solaris implements an mbuf model which was copied by FreeBSD with UMA, where Linux’s VMM interface has fundamental design errors which necessitate hacks like the OOM killer.
Enumerate these design errors please.
While Linux has kernel preemption, it substantially harms kernel throughput, whereas Solaris is able to achieve both the low latency of kernel preemption and high throughput by demarking non-preemption zones within the kernel where kernel throughput would be harmed.
Benchmarks please!
But don’t take my word for it.
Why would anyone take your word for anything?
Soon the forum will be beseiged by a hundred Linux zealots
What has the appearance (or otherwise) of Linux zealots got to do with anything?
(That’s what Fred Brooks said right? More programmers are better right? In his book the Factual Man-Month… oh wait, or was that the Mythical Man-Month…)
Sheer ignorance. Brooks’ described structured, commercial development. You’re attempting to make an apples-to-orange comaprison.
I contend in a large project like an OS kernel less is more… a cohesive experts group which has been with the code since its inception will, in general, produce better code than having random people from time to time cobble their own modifications onto the existing codebase, originating new features through trial and error (i.e. the round-robin I/O scheduler which has been exhibiting… unusual performance characteristcs) rather than through a professional research.
I contend you’re talking rubbish. First of all, a remarkable amount of peer review takes place on the LKML so getting code into the kernel is actually fairly hard, as it should be. By “trial and error”, I think you mean “experimentation”. This happens with closed source software as well (like Solaris), but behind closed doors.
Of course point out that Linux development has been mismanaged (Rik van Riel certainly contends that, http://linuxtoday.com/developer/2002011501620INKN)
Rik Van Riel is famous for spouting crap, and at that time he had a bone to grind. You contention that Linux development has been mismanaged is unsupported.
and you’ll get the typical OSS idealism which they think they can hold up to the practicality of the business world. Lacking real world experience with Solaris, they’ll link to a few poorly conducted, unscientific benchmarks.
Huh? You’ll certainly get people questioning your conclusions, and rightly demanding some evidence.
Solaris 10 is a revolutionary release, and ensures that Solaris will continue to be the performance leader for the near future
Revolutionary? Hah! Evolutionary, more like.
but rest assured that a zealously touted, mismanaged, cobbled together open source operating system which continues to get basic Solaris features 5 years after they were implemented in Solaris will continued to be touted by its brainwashed lovers as the performance leader simply because there is no way within their ideology that Linux could not have surpassed Solaris by this point… and no amount of benchmarks, accounts of real world experience, or code analysis will ever sway them from this viewpoint.
There don’t actually seem to be many people /making/ this claim, outside of your deranged imagination (and don’t you dare whine about this being a personal attack, your entire post is a personal attack). People are using Linux because it works, and it works extremely well. If zealotry had anything to do with it, FreeBSD and OSX would be kings of the world…
In the world where 5% better performance on DB test or great DTrace means better sales, Solaris would be on everyone’s desk. But it is not *this* world. For my computer, I expect: a) To handle my hardware, b) To have free sw for anything I want to do (if I am not doing rocket science), c) To have reasonable packaging and update scheme from internet. d) To be well-known and supported, so that I can get help on i-net when in trouble. For me (and many others) linux is better in these aspects than any other OS. Thats why we use it. If you like Solaris, gut luck. In none of those points (except in d, probably) Linux is the best OS outthere. But it is *overall good enough* and this is exact reason why people use it. For example I will not test Solaris 10, becouse I know it doesn’t support hw well. If I can’t install my printer, USB key, GeForce or sound card, I really don’t care for DTrace.
…and I guess next you’re going to say VMS is a horrible horrible server operating system because it can’t detect your USB keychain, or play DOOM3 correct?
Since when has anyone said Solaris 10 is a desktop machine? Since when do people choose servers based on printer support? I’m sure Amazon.com went with Linux over anything else from IBM, SUN, etc because Linux supported their lexmark printer… please.
Linux Good HW Support -> Large user base -> Lot of experts and “hardcore” users -> Chance that your company has these people inside -> Chance that your company moves to Linux soon
Blighty since you asked, here are a few links:
A few issues with the OOM:
http://lwn.net/Articles/111408/
http://seclists.org/lists/linux-kernel/2004/Dec/0110.html
Companies rely less on “hardware support” since they buy the hardware they need for what they want to run. For instance try reading the article here on Merc. its not a case of can I run the stock exchange on that P4 3.2ghz machine with a GForce 6800 I just got for christmas. Companies actually plan out their hardware/software infrastructure so that everything works.
Furthermore companies care about professional support mechanisms. I can promise you that large scale businesses couldn’t give a flying poop about “hardcore” users posting on message boards for support or that fact that maybe one of their IT admins runs linux at home for their USB keychain. Again see the Merc. article they expect things like flying out a certified tech to fix whatever the problem is.
Just once it would be nice have a linux desktop user not have the mentality that every corporation in the world can run their mission critical applications off a couple of desktop systems slapped together, a downloaded linux ISO and a few power users for consulting.
> Linux Good HW Support -> Large user base -> Lot of experts and “hardcore” users -> Chance that your company has these people inside -> Chance that your company moves to Linux soon
So called “good hw support” under Linux is a very questionable issue as the overwhelming majority of hw drivers out there are alpha quality at best or drivers simply failing to work with the new releases of the kernel (something you will never see under Solaris). If you factor out all Linux drivers with questionable quality, you’re going to end up with pretty much the same size of the hardware compatibility list as Solaris. And Solaris HCL is growing pretty fast, the list almost tripled in the last year alone, and that’s release quality drivers we’re talking about.
I don’t think you get it. Lets say that “for Enterprise class App, AIX and Solaris has better architecture and support then Linux”. I am Software developer of large Enterprise SW System. It used to work on Win NT only (Windows – another example of HW compatibility -> User Base -> Server Market chain). Since I have Linux at home, I modified C++ sources in such way that it can compile and run on it (only server, client is Java-based), so we will offer Linux version soon (it saves some money for Win Server on the costumer side). Can it run on AIX/Solaris/VMS? I don’t know, becouse none of our developers has those systems at home to test, and doesn’t look they will have it soon. Even our customers that has (very often) AIX-based datacenter, has no problem to buy additional (inexpensive) Intel Server(s) for our application. I cannot go to boss and say: How about spending $20K for AIX/Solaris + training, so that we can get +5% performance on superior kernels? He would probably think I am crazy. In two years performance will double, new version of our SW will save 30% of memory footprint so who cares? This is real life where company lives.
Very funny, how I see people argue that Linux won coverts because of “technical merits”, when in reality, it had nothing to do with that; talk to *ANY* early adopter, pre-linux hype, pre-dotcom days, and they’ll tell you that the main motivator was low cost coupled with relatively stable tagged onto the side of the box.
When Linux was making inroads, SUN was wishy washy with Solaris, and the only *real* UNIX on the x86 at the time was SCO UnixWare and OpenServer, both were over priced, especially for what many wanted their servers to do.
Linux was “good enough”(tm) solution for the time. It wasn’t the best, but it had the right mixture of cheap and reliable. The first adopters of Linux outside geek central were ISPs, they used them for email servers and the new server; a company run by IT geeks and maintained by IT geeks, meaning, running Linux and maintaining it was a non-issue.
Fast foward 9 years, and today, Linux is now facing competition, Solaris 10, it is now a real and viable competitor; even the manager from SUN admitted that had SUN gotten its act together sooner with their x86 line up, they most likely would have won the contract; atleast they’ve learnt, and you’ll find that in the future, Solaris 10 x86 taking more spots where Linux was once considered a viable alternative.
Do I hate Linux? nope, it has its place, just like FreeBSD, NetBSD and other operating systems, but lets not try to get into childish, “my operating system r000lz and your’s dr000lz”.
Gosh, is Sun paying all the pro-Sun shills here?
<<I’m sure it is going to be a very regretful move considering all new features and opportunities they are going to miss in Solaris 10 — Solaris 10 beats RedHat Linux on all counts from price and features to raw performance. I have to say it is a little stupid to forgo all the beautiful technology in Solaris 10 and just going with the Linux hype.>>
No hype in Linux – it has two critical attributes businesses have learned to value very highly. First, Linux is free: Free to download, copy, modify, share, distribute your own version, etc. When I see a free software community around Solaris like that of Linux, I’ll believe Solaris is equally free. When I see a Solaris distribution from a Sun competitor, I’ll believe Solaris is equally free.
Second, Linux is widely supported. Don’t like RedHat support? Switch to Novell. Or IBM. Or Mandrake. What happens if you don’t like Sun’s support?
More importantly from a business risk perspective – what happens if sun drops or cripples Solaris x86 support (AGAIN) to pump SPARC sales? Remember, fears of Sun bailing on x86 and screwing it’s customers aren’t just “what if” worries – Sun has ALREADY DONE IT ONCE!
Yes Solaris has nice features. I get a small performance win with the beta over Linux. But the 5%-10% isn’t nearly worth the risk premium in dealing with a single unreliable vendor like Sun when I have Linux, an alternative that has a demonstrated track record of reliability, stability, and steady performance improvements.
is SUN itself. Solaris could, and maybe is, the best OS out there, but it’s backed by SUN.
I have no use for SUN, don’t want them, don’t trust them, don’t need them, do want to hear from them, don’t like them. Mr. McNealy needs to take some serious business classes and get a dose of charm and civility.
And if SUN does truly OSS Solaris, any technology that is good enough in there will make it into GNU/Linux. If not, then…
Until then, the argument isn’t over which is better, Solaris or Linux. The argument is who is better, SUN or the FOSS communutity.
In other words, goodbye SUN.
(But I do wish you good luck so long as you play fair.)
When Sun execs claim that Solaris 10 has a good HCL on the x86 platform, are they pretending that in the last 12 months, they managed to write more drivers than during the past 5 years ?
> Second, Linux is widely supported. Don’t like RedHat support? Switch to Novell. Or IBM. Or Mandrake. What happens if you don’t like Sun’s support?
LOL. RedHat exec’s are already bragging about how prohibitive it has become to switch from one Linux distribution to another — 4 million dollars if I’m not mistaking. Sorry, but this theory of moving between Linux distribution is becoming more and more like a utopian dream — moving to another distribution means recertifying all application software, supportability concerns, etc., which in the end costs big bucks.
> No hype in Linux – it has two critical attributes businesses have learned to value very highly. First, Linux is free: Free to download, copy, modify, share, distribute your own version, etc.
The biggest reason Linux has become successful is because the free as in beer freedom in it — Linux became a runaway success because it was cheap compared to the alternatives not because of some altruistic freedoms. Business and enterprise sectors don’t give a flying duck about nerd politics propagated by GPL fanatics, all that is truly valued is purchase and support prices and TCO in the longterm. Solaris can address everything mentioned above in spades and be cheaper than Linux. Remember Solaris is targetted at the enterprise and not at the l33t hacker wannabe’s.
The article mentions using Linux over Solaris to get sub-second performance improvements, but doesn’t go into too much detail about how they determined that a hardware/OS refresh was going to solve their problems. Measuring that level of performance would be a challenge in itself, and again the article falls short. Did they eliminate any network latency issues, client performance issues or just simply “install Linux and run”? And lets not even discuss the applications involved, what tuning was done for them?
I don’t think Linux is a “cureall” for all ills, it is a tool just like Solaris, AIX, HP-UX, and Windows is. If you do not have people capable of examining and tuning both the OS, hardware, and applications then you are going to have problems.
<<<LOL. RedHat exec’s are already bragging about how prohibitive it has become to switch from one Linux distribution to another — 4 million dollars if I’m not mistaking. Sorry, but this theory of moving between Linux distribution is becoming more and more like a utopian dream — moving to another distribution means recertifying all application software, supportability concerns, etc., which in the end costs big bucks. >>
Heh. It’s only a “utopian dream” to incompetents. We did it (RedHat -> SuSE) without any issues beyond a week’s worth of startup script changes and testing. The key issue is that even if the $4M you attribute to Red Hat execs (it’s just SunShill FUD if you don’t provide a source) were true at least it is a option you DON’T EVEN HAVE if you’re foolish enough to base a business on Sun Solaris.
<<<Business and enterprise sectors don’t give a flying duck about nerd politics propagated by GPL fanatics, all that is truly valued is purchase and support prices and TCO in the longterm. Solaris can address everything mentioned above in spades and be cheaper than Linux. Remember Solaris is targetted at the enterprise and not at the l33t hacker wannabe’s>>>
Business care about propriatory lockin (how many sources for Solaris? 1). Business care about vendor strategic stability (Will Solaris support x86 next year or drop it when high margin SPARC sales slump). Business care about being able to get multiple vendors compete for their work (Solaris == only SUN, Linux == IBM, Red Hat, Novell, …)
‘leet hackers might care about a few extra percent of performance from Solaris, but a well-run enterprise business should be terrified of the high strategic risk that comes from depending on Sun Solaris.
> Heh. It’s only a “utopian dream” to incompetents. We did it (RedHat -> SuSE) without any issues beyond a week’s worth of startup script changes and testing. The key issue is that even if the $4M you attribute to Red Hat execs (it’s just SunShill FUD if you don’t provide a source) were true at least it is a option you DON’T EVEN HAVE if you’re foolish enough to base a business on Sun Solaris.
Err, obviously you don’t have much of an environment (and much of a clue for that matter either), do you? Changing platforms in the datacenter envrionment takes more than just changing a couple of scripts and considering it done. If you run serious enterprise class apps, just recertifying and possibly modifying the applications for a different distro will take a lot of resources and $$$, plus application support is not even close to being the same even with RedHat and SuSE — many applications are not even certified and not supported under SuSE for instance compared with RedHat. There is absolutely no pairity between Linux distributions as far as support for commercial applications in concerned — ISV’s standardize around RedHat, so there is practically no other Linux distro you can move on to and maintain supportability. It actually comes to the point where changing distros is as resource intensive as moving to an entirely different Unix platform alltogether (Solaris or AIX for example). The “freedom to change distributions” argument is becoming really blurred and really ineffective.
The “freedom to change distributions” argument is becoming really blurred and really ineffective.
This may be true, but…
GPL/GNU/OSS/FOSS is creating a HUGE pool of software and code from which a solution to your problem might exist (especially if it’s a common problem). An intelligent company is going to recognize the value of Linux engineers and sysadmins who can improvise and incorperate that technology into their system. That’s an incredibly duh statement, I know. But combine that with the need for actual skilled engineers (not those crappy certs) and the GPL/GNU/OSS/FOSS pool is free to drink from unlike SUN and MS’s pools and what you get is an opprotunity to restore high paying jobs for those interested in computer science and not those interested in a paycheck career. Otherwise you’re just certified to work on someone elses software and that is less desirable.
> But combine that with the need for actual skilled engineers (not those crappy certs) and the GPL/GNU/OSS/FOSS pool is free to drink from unlike SUN
Probably 99% of all GPL/GNU/OSS/FOSS software can run equally well under Solaris as on Linux. I don’t see much of a merit to this argument.
Why is it that everytime this sort of discussion pops up on OSNews I get the idea that the “Sun people” are somewhat irritated by that ‘childish’ os? People, deal with it. Do you like Sun? Use Sun. Do you like Linux? Use linux. What’s the problem? Are you intimidated by the success of Linux? Come on. There are bigger problems in the world than fighting about an OS….
<<Fact of the matter is that a huge number of businesses and organizations around the world have been comfortable relying on Sun for their mission critical operations and I absolutely do not forsee it changing any time soon — there are no strategic risks involved.>>
Reading from the SunShill handbook again? What does it tell you to say to all the Sun customers who were screwed by previous Sun strategic treachery on Solaris x86?
“Probably 99% of all GPL/GNU/OSS/FOSS software can run equally well under Solaris as on Linux. I don’t see much of a merit to this argument.”
What you just said is moot. I wasn’t talking about that at all. What I was saying is if you use SUN software, you are tied to SUN. If you use GPL/GNU/(true)OSS/FOSS, then you are tied to noone. So the question isn’t just the software installed software, the question is also the OS.
Answer this… why would I want to tie myself to SUN, developer or customer, when I can participate in GPL/GNU/(true)OSS/FOSS and the only obligation to the software I have is when I redistribute it?
Granted, SUN maybe better at somethings, but Linux will eventually be just as capable sooner or later without all the strings attatched. And if it’s truly something customers need, how long do you think it will be before the likes of IBM, Mandrake, Novell, Redhat, and the rest of the community to begin their own developement of similar features?
Freedom is not esoteric. Businesses do care about freedom, but sometimes freedom is too expensive so they opt for a compromise. But when the compromise becomes too onesided and benefit disappears, freedom doesn’t seem so expensive anymore.
With Linux, it’s possible to see a FORD distribution or a GMC distro. Say a NASA distro? Howbout a UCLA, CERN, Angola, Fred, Bob, Toaster distros? The point is, is if you share the binary, you have to share the code, otherwise you effectively OWN the software for all intents and purposes.
Heck, there is even a SUN distro… whoda thunk dat?
“Gosh, is Sun paying all the pro-Sun shills here?”
No, Solaris 10 is really the biggest release since Solaris first went 64-bit. The consensus on this is genuine.
” First, Linux is free: Free to download, copy, modify, share, distribute your own version, etc.”
Solaris 10 will be free to download, and the license for OpenSolaris supposedly is under review by OSI right now (thus, leading to free to copy, modify, share, distribute).
“Second, Linux is widely supported. Don’t like RedHat support? Switch to Novell. Or IBM. Or Mandrake. What happens if you don’t like Sun’s support?”
There are third-parties for Sun support, too, and Sun’s most recent pricing scheme falls right into line with RedHat, Novell, IBM, and Mandrake. There is no pure cost comparison that can give Solaris the disadvantage, anymore. Even Sun’s low-end hardware is priced to compete with Dell.
“what happens if sun drops or cripples Solaris x86 support (AGAIN) to pump SPARC sales?”
They won’t. (Are you a Dell or RedHat shill?) Opteron will be a perfect fit for people who want x86, open source, and Sun Solaris and support.
Look, what Sun is doing is taking the best that OSS offers and melding it with the best that Solaris offers. There are clearly a lot of great things about OSS, but there are also clearly things that Sun is ahead on. Sun’s customers will be better for it.
“Lingua Franca” is Latin for “the french language” and is iodiomatically used to refer to the “common language” of trade and diplomacy — a position now widley accepted to be held by English, though for centuries it had been French. Ironic that the Latin term for French now means English, no?
We see the same occurring here with operating systems. Linux is rapidly becoming the lingua franca of utility computing for several reasons: it’s a commoditiy (widely available, more or less free, useful for many things, people can grow it/bag it/sell it/buy it like a common garden vegetable, everybody recognizes it, etc.). As it has always been the case, features and technical superiority don’t sell something to the masses, only to the few. The computer industry is bathed in the blood of companies that learned that the hard way, and Microsoft exists today because of it.
What I’m getting at is that Sun’s Solaris 10 will not win or lose in the market place based on it’s features or performance (which is marginally better; incidentally, those that notice a difference should recompile their apps and kernel with the Intel compiler for Linux and then make the comparison). Linux isn’t going anywhere. It’s commoditized the OS for everywhere it’s landed in a niche where it’s even a marginal fit. But, for Solaris, Sun needs not only effective marketing, but also to honestly adress the issues that many of its current and former customers have with its products — and there’s quite a bit to overcome there.
I don’t think Sun’s current management is prepared to level with themselves or put forth the kind of effort that would be required to make Solaris 10 a roaring success. The engineers have done their part VERY well, but that’s not going to be enough. Solaris 10 is great, but it’s not going to sell itself, much less Sun hardware.
Also, I see Solaris and Linux being pushed in different directions. Sun still aims predominantly for the server space, while Linux is being pushed by some onto the desktop, others into embedded applications, others into the server space, etc. Solaris won’t fit the niches it doesn’t even make a move towards.
Personally, I use both Solaris and Linux quite a bit. Being a scientist, UNIX-like operating environments are everything I’ve ever wanted. I most frequently use Linux as my desktop OS (I’ve a OSX Mac here too, which is nice but doesn’t really match my style of using a GUI). I also do some work with Windows, which now feels very awkward and almost arcane to operate now. Despite my affection for Solaris, I’d probably still use Linux/*BSD/OSX on my personal machines — I’ve being using Sun software and hardware for a decade and I still have no sense that they care about their mid-size to small business customers or that they are truly committted to any product other than their SPARC processors.
I agree, very well written Bascule.
I work(ed) with many different OS’es (Linux, Solaris, Windows, Amiga, BeOS, etc etc..) and would like to add a few comments.
First, someone said Linux would gain since it used more at home and easily available for hardware. That is true, and Linux will be used for more home systems(desktop systems). Solaris is a server OS not a desktop OS. The nice thing about Solaris 10 X86 is you will be able to run Linux binaries on the OS. So create your applications on Linux, and when completed and need to be run on a server, just use Solaris (or hardware that Solaris will run on). So just because Linux is available for more hardware doesn’t mean it will be the logical choice to run on the server for compatibility.
Second, you got to realize a lot of performance (not talking just speed) is due to the hardware. Yes you can purchase cheap hardware, but you get what you pay for. Cheap hardware now in days will give you equal to better performance against a Sun system for example, but you don’t get the stability and probes that are available on a Sun system. How many cheap PCs people use as servers do you see that have as extensive diagnostics as compared for example to putting a Sun server into diag switch mode.
So what good is this extra diagnostics? This is one thing not talked about in the replies also available in Solaris 10, Self Healing. Here is an example: Solaris 10 monitors hardware, notices a memory module is erroring past the threshold, so it off lines just that memory module. In off lining that memory module, it notices a ZONE is using that address space. It determines it needs to restart that ZONE.
Solaris is able to do this since things like Zones and Self Healing you not just HACKED onto the OS like Linux does, but are tightly and smartly integrated.
Just wanted to add my 2 cents.
Yeah, i think it is partly much like Sun Microsystems has been overhyping quite a number of things in the news the past year or so, with cheerleader Jonathan Schwartz on the lead. With Solaris 10 and it open sourced nearing vaporware, and open source Java being an absolute attention drawer and lie. Not to say he hasn’t said anything interesting though, and its not as if Microsoft doesn’t do that (Longhorn *hint*).
I saw even a Sun.com person here admitting marketing spouts nonsens about features for Solaris 10 which are not in the betas and will not be in the final Solaris 10 release which are afaik: ZFS, the sourcecode. Instead they’ll be available later, after the release.
> What you just said is moot. I wasn’t talking about that at all. What I was saying is if you use SUN software, you are tied to SUN. If you use GPL/GNU/(true)OSS/FOSS, then you are tied to noone.
What a bunch of drivel. You’re not tied to Sun if you use Sun software — all Sun software with just a few exceptions runs on other architectures and operating systems, all of Sun middleware products at least run Linux, HP-UX, and Windows besides Solaris. If you don’t like Sun as a supplier of your hardware, there are hunderds of vendors to choose from to run the x86 side of things, for Sparc side of things there are always Fujistu-Siemens, Tatung, Themis just to name a few. In reality you get as much freedom with Solaris as with Linux especially if we talk enterprise applications — there isn’t much to choose from as only RedHat and SuSE are the only viable alternatives, only you still get a half-baked OS and crappy support from both RedHat and SuSE. Solaris is still a much better alternative on all counts to Linux.
RedHat exec’s are already bragging about how prohibitive it has become to switch from one Linux distribution to another — 4 million dollars if I’m not mistaking.
4 million for a company to move? Err, no.
Sorry, but this theory of moving between Linux distribution is becoming more and more like a utopian dream — moving to another distribution means recertifying all application software, supportability concerns, etc., which in the end costs big bucks.
It is easy. Buy servers in future from a different vendor – the open source software is identical and works in the same way. Red Hat will even be using Jonas as its J2EE server in the future, which makes it even easier. The dependency on Red Hat or anyone else quickly eradicates.
When the few customers that Sun attracts with this see the escalating support costs because Sun are propping up a free Solaris, they’ll be out of there fast. They’ve done this before.
If you actually knew anything about Solaris you wouldn’t be overly concerned about ZFS not being in the FCS release. Sun makes Maintenance Updates available to customers, and these updates add functionality, such as fssnap and improvements to Role Based Access Control in Solaris 8 MU’s.
Sun releases a version of Solaris every few months until the OS is End of Life (EOL). For example Solaris 8 was FCS released in June of 2000. The last shipping release other than hardware updates was February 2002. For me to bring my 6/00 installation to 2/02 I apply Maintenance Update 7.
I look forward to ZFS and not having to go through headaches carving up space for users.
> When the few customers that Sun attracts with this see the escalating support costs because Sun are propping up a free Solaris, they’ll be out of there fast. They’ve done this before.
What the hell are you bleating about? Solaris support is much cheaper than RedHat’s, SuSE or Windows — Solaris is the cheapest OS out there to support and license!
> It is easy. Buy servers in future from a different vendor – the open source software is identical and works in the same way. Red Hat will even be using Jonas as its J2EE server in the future, which makes it even easier. The dependency on Red Hat or anyone else quickly eradicates.
There is absolutely not lockin under Solaris either, you can run Jonas or whatever J2EE application server under Solaris just as easy as under RedHat. And moving from Solaris to Linux or from Linux to Solaris not much different than moving between different distributions — Solaris is already capable of running Linux binaries unmodified. This is a moot point as Solaris in practice offers you no less freedom and no more lockin than any commercialized Linux distribution.
The key issue is that even if the $4M you attribute to
Red Hat execs (it’s just SunShill FUD if you don’t provide
a source)
Perhaps you’ve forgotten the often-quoted source? It was
said at the Wachovia Securities 14th Annual Nantucket
Equity Conference by RHAT execs
http://blogs.sun.com/roller/page/jonathan/20040818#good_reading_lis…
Oh, that’s right, you [AC (IP: —.reston01.va.comcast.net)] don’t believe anything
that’s posted by Jonathon Schwartz.
How about this link then:
http://www.wsw.com/webcast/wa22/register.aspx?conf=wa22&page=rhat&u…
Schwartz: “The question which prompts the response starts
at 16:58 into the breakout (not the main) session.”
Business care about propriatory lockin (how many sources for Solaris?
Interesting that you say the above, when the vast majority are still running Windows XP and Office on their desktop.
Companies don’t give a toss about “lock” in, they worry about “best bang for the buck”, long term support stratergies, IHV and ISV support.
If businesses worried about “vendor lock in”, they would have started demanding an open standard file format long ago, the fact is, lock in isn’t a high priority for many businesses.
As for Solaris, how is it vendor lock in? it conforms to UNIX 98 specifications, don’t like Solaris, then move to HP-UX, AIX, SCO UnixWare or even Linux! As for “transferring”, please, don’t make out that you can twirl a server on a penny, want to move from Linux to Solaris or vise versa, try asking Oracle to give you a free license transfer; you’ll get a roaring laugh over the telephone. Think that IT employees can *suddenly* shift from one operating system to another without needing to retrained?
Please, if you are going to come up with an argument for Linux, please keep it in reality rather than this utopian fiction which you claim happens. Linux has just as much vendor lock in as the commercial UNIX world, the only thing Linux has going for it is “freedom, love and peace”; until we see fortune 500 companies being run by freedom loving beatniks, feel rest assured that the only things Linux will be doing is replacing low end servers and catering to niche markets which has specific needs.
Solaris 10 is not even released, it is overhyped.
> Solaris 10 is not even released, it is overhyped.
If I’m not mistaking Solaris 10 is due for the revenue release at the end of this month and according to some announcements from Sun there is already a number of customers running 10 in production. So, no it is not overhyped. Go grab some iso’s and try it for yourself, the product is available now.
You’re devotion to SUN speaks volumes.
Use them if you like, I know there is little to no chance that I’m going to. In the end, I believe that SUN and MS made a deal to split the market with SUN on the big hardware. I see little investment in Solaris other than to meet this goal. Because of this, it’s SUN that’s the problem, not Solaris. I don’t trust SUN. Solve why I should trust and use something like Solaris when the cry for them to release Java only highlighted their self-serving desire to keep it (and everyone else) under their thumb.
Until then, all your arguments about Solaris’ merits are moot. The only way I will use Solaris is if SUN manages to get a community going around Solaris that doen’t need SUN or its permission to use. I don’t care if SUN is around, I just choose not to depend on them, ever.
> In the end, I believe that SUN and MS made a deal to split the market with SUN on the big hardware.
May be you should also try on that proverbial tin foil hat. There is no conspiracy as Sun and MS are still the biggest rivals hating each other’s guts more than ever, it just so happened that Sun was successfull at extorting $2bn out of M$. Scott McNealy recently said that he likened the handshake with MS to tapping of boxing gloves between two boxers before they beat the living crap out of each other. If any sort of relationship to M$ revolts you, may be you should direct your feelings at IBM — the very company that has almost 5,000 full time staff dedicated to making Windows a success. You dislike Sun, but I bet you use OpenOffice on your Linux machine. The only reason Linux is making any inroads on the desktop is because of Sun — Sun spawned OpenOffice and made Gnome truly user-friendly. Even your beloved RedHat not so long ago (about year and a half) openly stated that Linux doesn’t belong on the desktop and that the desktop rightfully belongs to Windows. Then RedHat changed its stance in reaction to Sun releasing JDS. Sun is the biggest contributor to the open source and you’re biting the very hand that is feeding you.
For example I will not test Solaris 10, becouse I know it doesn’t support hw well. If I can’t install my printer, USB key, GeForce or sound card, I really don’t care for DTrace.
download the latest release and i’ll bet you it DOES support such hardware (printers require some config, but the support is there) perhaps except for your souncard, in which case you can use the free version of OSS for.
Solaris is superior to Linux for many kinds of problems. I think anybody who would argue that point is being foolish. But then again, you could argue that z/OS on an IBM Mainframe is an even superior solution as well as VMS on a cluster of Alpha’s. The point I’m trying to make is that Linux has never been, and may never be the “Best of Breed” technical solution. That is not its strength.
I have always been fond of a paraphrase of Winston Churchill. It goes along the lines of “Democracy stinks, but it’s better than everything else!” Linux is a lot like Democracy, for a couple reasons.
First, the technology either solves the problem or it doesn’t. When Solaris, z/OS, VMS and Linux are all adequate for a certain task, Linux really has a leg up because it’s cheap and starting to become comfortable. Windows succeeded mainly because it was “cheap and comfortable” to all the suits. All us geeks like jeans and t-shirts because they’re cheap and comfortable. Don’t underestimate Cheap and Comfortable. And don’t underestimate Comfortable, it will sometimes trump Cheap!
Secondly, Linux fulfills one of the original promises of Unix – openness. In the early days of computing vendors could keep their customers hostage, and thusly computing sucked. Unix was this little toy operating system (The CEO of Digital really did call it that) but it ran on practically everything and the cocky kids told everybody it would take over the world. The Mainframe and minicomputer vendors said “Bring it on!” and now they’re all gone. But Unix never really fulfilled that promise of openness. You may argue with me on what the real value of openness is, but even the Penguin Suit-wearing Scott McNealy thinks enough of openness that Solaris 10 will be open sourced in some fashion.
Solaris 10 will indeed be a “Kick Butt” operating system, but nevertheless I will probably not invest too much time on it. But if I had a multi-terabyte data warehouse to look after, or a large e-commerce site, I probably would. Personally, I cut my teeth on Dynix/ptx. That was an operating system that did a couple of things well, and you may not see its influence on AIX, but I assure you it’s there. It Kicked Butt, but I don’t really mourn its loss. I’m not saying Sun is going to disappear, but apparently its attitude of the day is “Hey Linux! BRING IT ON!”…where have we heard that before? Not all the Mainframe companies are extinct, heck IBM is going like gangbusters. But the IBM of today looks nothing like the IBM of the 70’s and 80’s and it was a hell of a bumpy ride to get here. If it is to survive, the Sun Microsystems of 2010 and 2020 will look nothing like the Sun of today.
In summary, what I’m trying to say is that Solaris 10, for all its many technical merits, may be irrelevant to what’s going on at the Chicago Mercantile Exchange, and in many other organizations. Is this something most people can agree on?
I have now reviewed the support issues we had during a feasibility study with RHAS: more than half the issues have not been resolved, in spite of more than 3 e-mail or phone interchanges in average, per issue.
I have to conclude that either RedHat support engineers are in average very ingorant and unintelligent, or have a vested interest in appearing so.
POST OF THE YEAR (even though the year is only 7 days old).
I completely agree and have been trying to say it that well for some time now. Anyone who has been in the industry before, the start, and now during the Linux explosion will also completely understand why you are so right. Otherwise they are either fooling themselves or are too new/young in the business side of IT that they can’t understand because they didn’t get to witness it and came in after it started (usually are the ones who are zealots).
Linux today and all of it’s merits, hype and acceptance is all 100% a product of such impecable timing like nothing I have ever seen before. Linux was nothing more than about 5 main distro’s from about ’97-99 and was mostly for the curious ‘geeks’ (i was one of them) to load on the extra server sitting around back at the office. The only deployments at the time were ISP’s, IF THAT! I worked for a very large IT consulting firm during those years and many just only have heard about it mentioned in tech magazines let alone were putting their jobs on the line in deploying or pitching it. I even remember *having* to go get Netware certified. And ’99 was ALL about Y2k. So much spending was done on IT and Y2k that year that when Y2k ended up being nothing but a farse, I think business as a whole got left with a sour taste in their mouths from us IT guys.
Hence the bubble burst in ’00. And it got worse and worse and what better of a product to run your business on than FREE and better yet “we have a guy in the IT dept that I heard is pretty good with Linux”. Cost cutting has been and continues to be the driving force behind Linux. “Why have 5 IT guys running our 3 Windows servers when I can get ONE to maintain one consolidated Linux server that cost me nothing for licensing no matter how much I need and can even use existing cheap x86 hardware! Sign me up”!
How many IT/Tech/programmers/etc lost their jobs since mid 2000? I think the economy has had much more of a driving force of Linux than its technical side. Over the past 6 years I continue to see nothing but catchup/emulate attempts by linux. It REPLACES expensive proprietary OS and software with basically free onea that can do the same things that business needs. Web/email, DB, hell it even can pretend to be a MS smb file server, and now with its own attempt to even be an Active Directory smb server. Open Office is nothing but (and of course LOOKS) attempts to be like MS Office. KDE looks like Windows. Evolution is Outlook. Etc. Shall I go on?
Linux isn’t going away but I can see it being negated and topics of news with linux being replaced when a) the economy starts to get back to boom days it was in late 90’s and b) proprietary OS makers get done adjusting to the blow Linux has delivered. Sun FINALLY looks to have woken up, and MS has dropped pricing. Even Apple has benefited from the Linux explosion. I am grateful for the changes linux has and will continue to bring in my proprietary and even other Open source OS’s (BSD’s). But like 80’s music, it had it’s time and now it is just worthy of flashback night at your local clubs…
PS – Novell will be hurt the hardest as they have nothing else but linux to save them as Netware was already dying without Linux. And when linux slows down, Novell = bye bye