If you were expecting a sneak peak of an open source Solaris or to buy a commercial version next week when it launches, don’t hold your breath. Although Sun Microsystems said it is on track to officially launch Solaris 10, the next generation server operating system, on Nov. 15 at its Network Computing 04Q4 event in San Jose, the products will actually take a bit longer.
Open Source Solaris Stalls
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
Maybe nobody writes exploits for Solaris cause they’ve historicly costed an arm and a leg and 3rd world “h4x0rz” don’t write code for sparc or whatever. We’ll see how secure it is when x86 is all over the place.
Anyway, this security topic has been beat to death, anything can be secure or broken.
Here login as root on fedora and see if you can break it.
http://www.coker.com.au/selinux/play.html
or Debian:
http://selinux.simplyaquatics.com/
or Gentoo:
http://selinux.dev.gentoo.org/
“Trusted Solaris” is just as good but SeLinux is being added to mainline not just ‘hardend’ distro’s.
The study was made by mi2g (http://vmyths.com/resource.cfm?id=64&page=1). I wouldn’t recommend making any important decisions based on their studies…
ObOpenSourceSolaris:
It will be interesting to see if Sun actually goes through with open-sourcing Solaris, and in that case how much of it will be opened. After all, Solaris contains a lot of third-party code which can’t be opened (as the article mentions).
So, will they just release a bunch of incomplete, non-compiling code (sort of like Netscape did with the first Mozilla code), or are they working on replacing the stuff that can’t be opened?
“Both HP and IBM are abandoning HP-UX and AIX,” Schwartz said. “What else do they have in the Unix market?”
You heard it here. HP-UX and AIX are both dying, pushing up daisies. I think netcraft has also confirmed it.
More than 65% of all break-ins occur on Linux systems. 25% Windows, etc…
Oh a report by mi2g. How informative.
“There is no binary for Linux, so they may have a dependency that any binary from Red Hat or SUSE is flawed,”
Excuse me?
The cross-partnerships are no bother to Schwartz, who said he even expects HP (Quote, Chart), and perhaps even IBM (Quote, Chart), to adopt an open source version of Solaris.
“Both HP and IBM are abandoning HP-UX and AIX,” Schwartz said. “What else do they have in the Unix market?”
Errr. Linux?
Solaris is so damn optimized and so damn stable it puts Linux to shame.
Ah, so THAT is why Sun don’t allow you to publish any benchmarks – they want to keep this wonderful Solaris goodness a secret so it doesn’t get spoiled by the unwashed masses.
I have another scenario for you: Linux puts Solaris to shame. How do you like them apples?
Right on!
“Ah, so THAT is why Sun don’t allow you to publish any benchmarks – they want to keep this wonderful Solaris goodness a secret so it doesn’t get spoiled by the unwashed masses.
I have another scenario for you: Linux puts Solaris to shame. How do you like them apples? ”
Before comparing your precious Linux to an enterprise class operating system like Solaris, why don’t you first try to beat the lowly Microsoft Windows in benchmarks? The web is litterred with hundreds of benchmarks showing Windows kill Linux in performance. Do a google search, but here is some on Microsoft’s website:
http://www.microsoft.com/windowsserversystem/facts/analyses/msvssam…
http://www.microsoft.com/windowsserversystem/facts/analyses/webbenc…
http://www.microsoft.com/windowsserversystem/facts/analyses/wfsthru…
Now before wetting your panties and going on a rampage, I wasn’t trying to say that either Windows or Linux were better. I was just making a simple point: anyone can make benchmarks say what they want them to say. If I was inclined to, and if I was given proper permission, I would even be able to show benchmarks of OpenBSD outscaling Solaris as you add more processors. Remember, there are lies, damned lies, and benchmarks. This is the primary reason why companies do not let you benchmark their products without proper permission.
You quoted Microsoft’s studies?
That’s a bad joke.
Please read my entire post. I was simply making a point as to why companies put restrictions on benchmarking. I think you will agree that a) benchmarks are useless and can be made to say just about anything and b) companies/people can use them to slander other people’s product (and I’m not only talking about Linux zealots slandering Solaris).
So you don’t belive in benchmarks, fair enough.
———-
Slash:
“Solaris is so damn optimized and so damn stable it puts Linux to shame.”
Anon:
“I have another scenario for you: Linux puts Solaris to shame.”
———-
Both comments provide no evidence to back up their claims, yet you come out on the offensive. Why? Unless you’re willing except some sort of technical comparisons, how can you attack people’s posts and call them “linux zealots”?
Stop being a hypocrite.
Whoops my bad, that was AndrewZ’s post
Sorry slash.
I can see anything with these glasses
Actually, what I said was a sarcastic joke along the lines of how a zealot makes stuff up and of course has no evidence to back up these claims, but starts preaching them like gospel.
Ironically, slash, you even went one further and quoted a Microsoft study putting a Microsoft product in a good light – something which everybody *knows* is suspect (if you say otherwise you’re very naive).
spank_da_monkey: at least you’re on the right track even though you interpreted it the humourless way. Look, now that Andy’s post has been removed I’d be quite happy for mine to go too.
have been using solaris9 on sparc, openbsd was faster
have been using JDS on x86, linux was faster
have downloaded solaris10, which one was it ? workstation or server or both ? dont care anynore, when i visit sun.com i get confused of theirs solaris with that, JDS with Linux, JDS with solaris, solaris 10 for …….
one day free, yesterday/nextday non-free
one day opensource, yesterday/nextday closedsource
one day solaris 10 with dit and dat , yesterday/nextday with dit and dat
sun.com is worse than Mandrake and Novell on their presentation on their products
Mandrake with their silver? memberships and their powerpacks
Go Fedora (maybe Debian also) & *BSDś
sorry, for my bad English
NetBSD 2.0 is soon out
I downloaded b69 (beta 7) from my ultra 5.
I’m absolutely stunned that its the same old *crap* installer. Surely these guys are joking (sun that is). Crapping on about “Dtrace”, “ZFS”, Firestack (TCP rewrite) etc. HOw about the ability to setup a mirrored disk environment at BOOT TIME. Linux has been doing this for years ! Not to mention the fact that an ssh login takes an eternity under solaris 10 and flies on openbsd 3.6.
Solaris has a long long long way to go before it gets anywhere near linux for general purpose computing…..and linux is catching so fast in the enterprise space that it wont be long before its beating Solaris in that space too.
Roll on Fedora Core 3.
Perhaps if you had paid attention at the boot prompt you
would have noticed that there are now four separate options
for how to boot and do the install.
Perhaps if you read the release notes (now there’s an idea!)
you would have seen that the old installer is no longer the
default, you actually have to choose to use it.
If you were actually interested in what the new features
which are now available in Solaris Express can do for you
then you wouldn’t be talking about “crapping on.”
If you’re going to post comparisons with other OSen,
at least provide some data so we can determine whether
you have a clue to base your comparisons on.
Oh, and it’s “FireEngine” btw, and the performance claims
are justified — I’ve seen this with my own eyes.
ZFS isn’t in yet — I’ve heard that the engineering team
has this crazy idea that they shouldn’t release a product
without making sure that it performs to expectations and
doesn’t corrupt your data. You might not care about that
but my boss and I certainly do.
DTrace — well, if you haven’t ever wanted to find out in
real time where your performance bottlenecks are/were —
without decreasing your system performance — then clearly
you aren’t going to be interested in DTrace. Most people
I know who work with solaris (or in a commercial environment
in general) are interested, because they want to squeeze
the last drop of performance from their systems. DTrace
actively assists them with doing just that.
I suspect that when you talk about “general
purpose computing” that you actually mean
“on my desktop” — that’s not where enterprise
computing is targeted. Enterprise computing
is targeted at companies which pay money for
solid engineering, solid support and binary
compatibility so they don’t have to recompile
their apps every time a new version of the
kernel is released.
Hey @McBofh, I totally agree with you!!!
Why everytime I read the words Solaris and Enterprise computing, there are a bunch of so called GNU/Linux Zealots whining about slowaris and stuff like that???
C’mon, I guess its envy you know.. every single admin who had the ability to maintain a proper Solaris enviroment would agree with me and McBofh.
Lets all be clever enough to not turn OSNews into a new Slashdot..
Regards and respect,
jay
Hi,
Finally, yes, thankyou McBofh for injecting some commonsense and logic into this flame-fest =).
Seriously, with the way these ‘zealots’ go on about “Solaris is, like, totally crap for my general-purpose computing”, whatever the heck that’s meant to mean. No, Solaris wasn’t designed so you could flawlessly watch your p0rn films and play Tetrinet – it was made for, *gasp*, real work – databases, web servers, mail servers and other things beyond your toying around with linux (which personally I like a lot).
And what do you mean whining about the installer, of all things – wouldn’t most enterprises and *serious* admins use a Flash archive anyway?
Can’t wait for ZFS myself…
bye,
Victor
Well my first post on this topic was deleted, because it was apparently seen as trolling. My apologies if I didn’t provide enough substance to back up my claims that Solaris has optimizations well beyond Linux and Windows.
Firstly, Solaris includes an optimizing C compiler which is superior to GCC. This is one way it is better optimized. The list of options on this compiler is many, many pages long.
Secondly, Solaris may very well be the best scaling OS for SMP operations. It sports a very good clean context switching kernel, and very fast shared memory communications.
I’m surprised that Sun hasn’t pushed Solaris as a competitor to Beowulf clusters. Maybe we will see this in the coming months.
Thirdly, you can do system and kernel optimization tuning on the fly and under load. Wow. No reboots before you see how changes affect performance.
So I’d like to say again, after having provided some grounds for my claim that Solaris was designed from day 1 for large scale enterprise performance: I’m glad to see that Solaris will be around for a while, and will hopefully outlast the SPARC architecture it rode in on.
Regards,
AndrewZ
where can i get beta 7 from ?? everytime i try to download i get beta63 beta 6 ?
So what if Solaris is better then Linux, Linux will so catch up. It has more people working on it, more people using it, and more money going into it, and it will soon engulf anything else. You Unix Zealots are no better then Linux zealots.
“It has more people working on it, more people using it, and more money going into it, and it will soon engulf anything else”
First of all, what you describe seems to be an extremely drab world. Why would anyone want a world where there is only Linux. Isn’t it like wishing for a world with only Ford cars.
Secondly, even if/when Linux does catch up to Solaris, it doesn’t mean Solaris will disappear. Just as Solaris, AIX, HPUX, etc. all coexisted, there is room for everyone else to exist with Linux too. It is impossible for Linux to kill Solaris in performance. Solaris is a solid Operating System that uses the vast majority of hardware that is available to it extremely efficiently. So even if Linux does become much better than Solaris, the difference in performance between Solaris and Linux will be negligable.
BTW, I am writing this on a P4 Windows 2000 Professional Workstation running Mozilla FireFox. It is using a mere 4% CPU utilization. I am content with my system but I would be just as content with my system if I was using Linux, BSD, or Solaris. Firefox and OpenOffice run on all of the above.
First of all we are not talking about desktops, Windows is locked in good in that market. The next thing is that AIX, HP-UX are disappearing even Schwartz said that they other day. Why you ask? Because both HP and IBM are going to Linux. It is not impossible for Linux to kill Solaris in performace because Solaris runs on very limited hardware. Have you ever looked at the HCL? and then looked at the one for Linux? Linux is surperior because in this because your not stuck with so limited hardware. Linux does a real good job in performance with the vast amount of hardware that is suported by it. One of the reason’s your machine is settin at 4% cpu utilization is the fact that your using OPEN SOURCE SOFTWARE.
More linux users filling yet another thread with garbage. Why can’t we have a nice thread that is about…. solaris? You know that thing mentioned in the title.
The Linux zealots don’t get it, why would would you want to run on “off the shelf” hardware trying to simulate a larger machine. Some things are just done better on expensive proprietary hardware. Seymour Cray had it right “If you were plowing a field, which would you rather use? Two strong oxen or 1024 chickens?”
When Linux runs on SunFire Midrange and High-End servers, supports Domains (hardware) and Zones (software), and has the full range of tools that normally ships with Solaris, THEN I will take Linux seriously. I really don’t see HP or IBM “selling the farm” to go to Linux either, they are using it to supplement their existing offerings. Yes you can buy a pSeries running RedHat or SuSe Linux, but does either OS support LPAR’s or DLPAR’s (NO), that is why you run AIX on that class of hardware.
http://www-1.ibm.com/servers/eserver/pseries/hardware/highend/595.h…
[i]”There are limitations to what we can do,” Weinberg said. “We don’t own all the IP rights to Solaris. There are things we have from third parties we don’t have rights to yet, or third parties don’t want us to expose source code.”<i/>
Hence this claim of wanting to open-source code that has been closed from the begining is poorly thought out. I doubt SUN can do it without seriously breaking the OS.
As far as linux compaired to Solaris in my opinion its night and day. Linux isn’t even close to the quality of code present in Solaris. The more i work with linux the more i start to dislike it…ever try to set the duplex on a NIC during boot in linux with a card not mii compatible? its ugly. Solaris ndd command in a startup script and your golden.
Linux isn’t even close to the quality of code present in Solaris. The more i work with linux the more i start to dislike it…ever try to set the duplex on a NIC during boot in linux with a card not mii compatible? its ugly. Solaris ndd command in a startup script and your golden.
—-
what are you smoking. mii is not dependant on any driver. linux will work on more lan cards than any other operating system esp on x86 architecture. solaris is no where near the quality and breadth
Not to mention the fact that an ssh login takes an eternity under solaris 10 and flies on openbsd 3.6.
Your DNS config is broken.
Well my first post on this topic was deleted, because it was apparently seen as trolling. My apologies if I didn’t provide enough substance to back up my claims that Solaris has optimizations well beyond Linux and Windows.
You’ve still got a long way to go.
Firstly, Solaris includes an optimizing C compiler which is superior to GCC. This is one way it is better optimized. The list of options on this compiler is many, many pages long.
You may be completely right, but you have provided zero evidence. If you think the number of options on a compiler being evidence to how good it is, then you are seriously misguided.
Secondly, Solaris may very well be the best scaling OS for SMP operations. It sports a very good clean context switching kernel, and very fast shared memory communications.
It may be indeed the best scaling OS. It also may not be, with all the evidence you’ve provided. I’d say IRIX will definitely beat it, AIX might too, and Linux might be able to as well (the first two are difficult to verify as they run on different architectures).
Context switching, I assume you mean preemptive multitasking. No big deal, every decent operating system has been this way for many many years.
Shared memory, ditto. It’s no big deal.
I’m surprised that Sun hasn’t pushed Solaris as a competitor to Beowulf clusters. Maybe we will see this in the coming months.
Solaris is an operating system. Clusters are computer hardware, so I don’t think Solaris could compete with them.
Thirdly, you can do system and kernel optimization tuning on the fly and under load. Wow. No reboots before you see how changes affect performance.
Wow, so can most operating systems (except maybe NT).
So I’d like to say again, after having provided some grounds for my claim that Solaris was designed from day 1 for large scale enterprise performance: I’m glad to see that Solaris will be around for a while, and will hopefully outlast the SPARC architecture it rode in on.
Well yeah *after* having provided grounds for your claims, sure. You haven’t yet.
The Linux zealots don’t get it, why would would you want to run on “off the shelf” hardware trying to simulate a larger machine. Some things are just done better on expensive proprietary hardware. Seymour Cray had it right “If you were plowing a field, which would you rather use? Two strong oxen or 1024 chickens?”
When Linux runs on SunFire Midrange and High-End servers, supports Domains (hardware) and Zones (software), and has the full range of tools that normally ships with Solaris,
Pfth. Sorry, Sun hardware is nothing special. See for example http://www.spec.org/jbb2000/results/, where Linux on a 16-core p570 trashes everything Sun put up except for a 100+ CPU system that just edges it out.
Hey I’ll take Linux and 512 big oxen in an Altix (soon 2048) (Itanium 2 is basically double the fastest Sun SPARC on FP performance, and faster on integer) – rather than your 144 little weaking oxen.
Sun’s interconnect used to be pretty cool back in the day, I’ll give you that. It now resembles a dinosaur, especially with all those SPARC chips plugged into it.
THEN I will take Linux seriously. I really don’t see HP or IBM “selling the farm” to go to Linux either, they are using it to supplement their existing offerings. Yes you can buy a pSeries running RedHat or SuSe Linux, but does either OS support LPAR’s or DLPAR’s (NO), that is why you run AIX on that class of hardware.
Err yeah sure Linux can run under the hypervisor. Newer kernels can do the full range of CPU and PCI hot plug and unplug, so I think DLPAR would be workable too. Got any more FUD?
I didn’t say I was going to write a mathematical proof for you and I didn’t. But none-the-less I stand by my assertions and trust the benchmarks will back up my personal experiences.
And hey, Irix was cool. As was the original Amiga OS at the time. But forget about a future support plan. You’d be a fool to run your company’s critical applications on it with no future.
Quite frankly I’m kind of surprised that I got modded down here and you didn’t.
– AndrewZ
I didn’t say I was going to write a mathematical proof for you and I didn’t. But none-the-less I stand by my assertions and trust the benchmarks will back up my personal experiences.
Not only did you not “write a mathematical proof”, but you didn’t provide any evidence whatsoever. In fact, your claim that Solaris may be the most scalable OS is pretty laughable when you look at IRIX (I’ve even seen AIX handily outscale Solaris, although that may be more to do with the underlying hardware, so I won’t make any claims about that).
And hey, Irix was cool. As was the original Amiga OS at the time. But forget about a future support plan. You’d be a fool to run your company’s critical applications on it with no future.
So… change the subject? Or admit Solaris is not the most scalable?
Quite frankly I’m kind of surprised that I got modded down here and you didn’t.
Why? Because I said something you don’t like? As opposed to you making claims with zero facts to back them up?
Sorry, I’m going to continue to pick people on these points. If I were to go ahead and say Linux is better at xxx than Solaris without any evidence, then I definitely would get shouted down too.
Mate, what are *you* smoking?
Yes, the Solaris compiler is better than GCC, and it’s well known and accepted – as is the ICC compiler by intel. In fact, there was an article in Linux Journal in the last month or so about compiling the linux kernel using ICC instead of GCC for the performance gain.
There’s some comments on ICC here:
http://www.linuxjournal.com/node/6766
(Pro: 30% faster than GCC, Bad: limited compat. with Gcc =_)
and some info on GCC versus Sun Compiler (Studio?)
http://www.osnews.com/story.php?news_id=5830%20target=
bye,
Victor
OK what are you smoking? WTF bring ICC into it at all? That seems to be the backbone of your argument.
Actually, Sun’s compiler versus GCC on SPARC is surprisingly unimpressive in those tests you linked to. It is a tiny bit faster.
Sorry, this says *zero* about Linux versus Solaris, the operating systems. You can just as well run ICC on Linux and compile Linux with ICC. Conversely, you can compile plenty of Solaris apps with GCC. GCC != Linux, mate.
My concern is real world benchmarks, day to day performance. And of course newer hardware with better technolgy is going to be “faster”, so what!
First hot plug PCI is not the same as Dynamic Reconfiguration (Sun) or IBM’s or HP’s equivalent. The OS has to be built for that level of hardware support. With DR, I can remove a system board and add memory while the machine is hot! This a so beyond hot plug PCI. So lets not compare “apples and oranges”.
And while we are talking about “apples and orages”, why are you comparing MPP (Massively Parallel Processing) to SMP (Symemetrical Multi Processing)? MPP boxes are built to run one application only and scale across a number of CPU’s (NEC’s Earth Simulator and the NSA’s Cray SV2). Noe of the machines I just mentioned are meant for general purpose computing. So try to run Oracle and SAP on your cluster and send me the results. Better yet, read what SGI says about their own machine:
http://www.sgi.com/products/servers/altix/
Particularly this sentence:
Access hundreds of 64 bit technical applications with Industry standard Linux
Technical applications are built from the “ground up” for a particular platform.
So much for your “FUD”
If IBM or HP announces some great new server, you can be sure that sunw will follow the announcement, with sunw’s own announcement about how sunw is now “considering” open-sourcing solaris.
Sun can’t open-source solaris, much of the code is proprietary.
Besides, sunw doesn’t want to open-source solaris. I marvel at how often people keep falling for the same old tired PR stunt. Is it really news anymore when sunw announces they are “considering” open-sourcing solaris? Wake me up when sunw really does something.
That is why Sun is going for the XOrg X Server and will probably replace Xsun with it. That is more than likely the first step in gleaning SVR4 code that would prevent Sun from releasing a “Open Source” version of Solaris and getting sued by SCO and Novell.
I don’t exactly see IBM or HP “stepping up to the plate” to “Open Source” their primary OS (AIX and HP-UX).
What I don’t understand is why Sun attempts to do something good and everybody has a complaint!
Sun can’t open-source solaris, much of the code is proprietary.
This is why the effort has taken more time than initially expected. The title of this article is really a misnomer. In particular, Sun is making sure that they get all of their legal ducks in a row, and are re-implementing many of the proprietary parts of Solaris that were licensed from 3rd parties
Sun can open-source Solaris, it just requires great scrutiny about their legal obligations. They need to make sure they know what code needs to be re-written, junked, or offered some other way.
First hot plug PCI is not the same as Dynamic Reconfiguration (Sun) or IBM’s or HP’s equivalent. The OS has to be built for that level of hardware support. With DR, I can remove a system board and add memory while the machine is hot! This a so beyond hot plug PCI. So lets not compare “apples and oranges”.
Do you know what you are talking about? In Linux you can do that as well. You can {un}plug CPUs, PCI devices, PCI bridges, while the system is running. You can’t hot plug memory yet, although that’s being worked on IIRC by Fujitsu and NEC.
And while we are talking about “apples and orages”, why are you comparing MPP (Massively Parallel Processing) to SMP (Symemetrical Multi Processing)? MPP boxes are built to run one application only and scale across a number of CPU’s (NEC’s Earth Simulator and the NSA’s Cray SV2). Noe of the machines I just mentioned are meant for general purpose computing.
Sorry chap, Sun’s SunFire systems ARE NUMA. They have 3 tiered memory access. Latency and bandwidth ratios from local to remote are not really any better than Altixes fo the same size.
So try to run Oracle and SAP on your cluster and send me the results.
Altixes are not clusters. Your deep understanding of all the issues is shining through. SGI has run a 2048 CPU cache coherent single global memory Altix at NASA. They have numerous 512 CPU systems in production.
Better yet, read what SGI says about their own machine:
http://www.sgi.com/products/servers/altix/
Particularly this sentence:
Access hundreds of 64 bit technical applications with Industry standard Linux
OK.
Technical applications are built from the “ground up” for a particular platform.
This is the best argument you can offer? That Altixes can run “technical applications” is your basis that they’re no good?
So much for your “FUD”
Ahem. Please go away.