There’s a benchmark comparing Solaris to Red Hat at Sun’s web site. Solaris 10 features a new TCP/IP stack architecture, project FireEngine. Sunay Tripathi posted some performance data on his blog as well that people might be interested in. He will also be posting details about the new architecture and how it allows Solaris 10 to perform exceptionally well on 1-2 CPU and also scale linearly across large number of CPUs. With the low end x86 platform moving soon to 8 CPU (and AMD’s dual core, 8 CPU), scaling is something that can’t be ignored anymore.
Solaris Performance Benchmarks
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
sure sun is going to spiffy up the look of solaris compared to anything else, who in their right mind is going to publish data that displays unfavorable results about your own product…
microsoft been doing this for years, sun finally figure that out? the quality of articles to comment about here at osnews is getting really poor, maybe osnews should publish a report showing how much better osnews is than say slashdot or any of the other comment and rant collecting sites around the net…
nothing to see here folks, move along…
Wow, Sun’s telling me that Solaris is faster than Red Hat? What a shock.
Seriously…some of these benchmarks are cooked…look at the JVMs and application servers used on each. Everyone knows that Java 1.5 is much, much faster than Java 1.4.2. So the only benchmark that is truly valid for comparing OS performance is the Apache one.
Also, is it stock Solaris vs. stock Red Hat or performance tweeked Solaris vs. stock Red Hat?
Not to belittle Solaris 10…it looks like a truly great OS, but I’ll believe these numbers when an independent source achieves the same results.
Thats not what is being said. The other posts are simply saying that the benchmark is publish by Sun on a Sun product. Obviously it is going to show its product on a favorable way. Questioning the integrity of a benchmark is not the same as saying linux is getting beat unfairly.
Yes, the first post was whining “Oh…I guess they finally learned how to do what Microsoft has been doing for years.”
And yes, this happens with independant benchmarks as well… Then the conspiracy theories fly about how Microsoft or some other such company probably paid for the benchmark and for the results to be biased.
The simple fact is that this benchmark does not have the same vaidlity as a study published from a reputable third party source. The data may be totally accurate but its coming from one of the parties involved. Its like in politics when a canidate talks about himself vs another canidate.
Actually, what I find funny is how predictable you Linux zealots are.
Doesn’t matter how good the benchmark is… You Linux zealots will always whine and pout and stomp your feet about how a benchmark is rigged anyime it shows that another OS outperforms Linux.
Do the world a favor? Admit that Linux has some flaws and then improve on it instead of just trying to claim that it is better than any other OS out there and complaining about any benchmark that says otherwise.
I don’t see anybody whining because Linux got beat…I don’t doubt that other OSes are faster than Linux (in fact I bet FreeBSD 5.3 final will be)…I see people calling a spade a spade. This benchmark is as useful as Apple’s famous benchmarks.
When someone independent performs an objective benchmark and gives us the full extent of configuration details I’ll take some stock in the results.
However, you must admit that the fact that Sun is even comparing Solaris to Red Hat Linux does not bode well for Sun. It’s kind of like admiting that little brother Linux is now almost in Sun’s league.
Just some food for thought.
“The ISP now sends 25,000 messages per second on the same AMD Opteron processor-based hardware and has abandoned Linux.”
So much for Sun being a friend of linux. The din saying Sun still loves Linux, just not Red Hat, is getting shrill and staticy.
“However, you must admit that the fact that Sun is even comparing Solaris to Red Hat Linux does not bode well for Sun. It’s kind of like admiting that little brother Linux is now almost in Sun’s league.”
Sun can’t keep trying to hide from what customers already know, which is that RH Enterprise Edition is an alternative to Solaris. For Sun to simply deny this would do them no good at all. So they have to show why Solaris is better than Linux.
I really don’t get why Sun are only gunning for Red Hat here, isn’t there another server OS that runs on commodity hardware they should be competing against? I think it’s called something like “Windows”.
Or are Sun now completely Microsoft’s b*tch?
Disappointing, truly.
Yeah, you need the 3rd party benchmarking.
Now when the third party benchmarks are done and Solaris 10 comes out on top, how many rabid fanboys are going to come out and say that the 3rd party is on the take?
RHEL 3 is based on a 2 or 3yearold kernel release(2.4.21). It doesn’t perform exceptionally well, but it’s stable. Solaris 10 on the other hand doesn’t appear to be out yet. When rhel3 cameout opterons just came out, so obviously it can’t be as optimized as solaris10.
Maybe they should be benching against rhel4 beta since that would be more on par.
They’re outperforming redhat AS 3.0 by 15-35% in the 2-way field…ok. Those would be nice numbers, until you realize that redhat as 3.0 is based in the 2,4 kernel. 2.6 already outperforms 2.4 by a big margin, I’d have expected more from solaris….
Sun can’t keep trying to hide from what customers already know, which is that RH Enterprise Edition is an alternative to Solaris. For Sun to simply deny this would do them no good at all. So they have to show why Solaris is better than Linux.
The point is that Sun is used to having to compare themselves to HP-UX, IRIX, and AIX. The fact that they see Linux as a threat at all is a major win for companies like Red Hat. It’s not hard to see that Solaris is about to be Netscaped by Linux. In other words, relinquished to niche markets like extremely high-end servers and organizations who are too locked-in to see much benefit from such a switch.
How about RHEL 4 Beta 1 vs Solaris 10? Would make a lot more sense considering Solaris 10 isnt even avaliable yet.
Redhat seems to me a much bigger competitor than windows for sun. Linux is unix-like, windows is not, you can easily switch from solaris to linux (much easier to migrate your software, no need to completely re-educate your programmers/sysadmins)
>2.6 already outperforms 2.4 by a big margin, I’d have expected more from solaris….
A big margin ? The test says Solaris to be 15-35% faster. I havn’t seen terribly many benchmarks saying 2.6 is overall much than 10-20% faster than 2.4
http://www.google.com/search?q=linux+kernel+2.4+vs+2.6+benchmark&so…
http://www.2cpu.com/articles/98_1.html
http://kerneltrap.org/node/view/903
…
> …saying 2.6 is overall much than 10-20% faster than 2.4
Oh, almost forgot. RHEL kernels already contain backports of a whole lot of scalability enhancements from 2.6.
“It’s not hard to see that Solaris is about to be Netscaped by Linux. In other words, relinquished to niche markets like extremely high-end servers and organizations who are too locked-in to see much benefit from such a switch.”
I disagree. For several reasons:
1. Sun is open sourcing Solaris.
2. Sun is a company that most major corporations have dealt with before, and are comfortable doing business with. Red Hat is still considered a startup that is fragile and could be bankrupted by a single major lawsuit.
3. Sun is offering a true desktop solution for Unix, whereas Red Hat is not.
4. Solaris is significantly cheaper then RH Enterprise.
5. Sun already has case studies of customers who they have migrated away from Linux, and the customers are much happier with Solaris than they were with Linux.
Some of Linux’s not so impressive performance numbers on Opteron are likely due to GCC and not Linux itself.
GCC is not known for producing very efficient code on 64 bit systems. Sun is likely using a propreitary compiler to build Solaris 10 that is probably much more efficient at producing 64 bit code.
. Sun is open sourcing Solaris.
—–
tell me after they do it. they have been telling that for quite sometime. what about sco?
“2. Sun is a company that most major corporations have dealt with before, and are comfortable doing business with. Red Hat is still considered a startup that is fragile and could be bankrupted by a single major lawsuit. ”
what world are you in. oracle just shifted from sun solaris to redhat linux completely for inhouse development. various companies trust redhat now. they arent locked into one vendor with linux either
“3. Sun is offering a true desktop solution for Unix, whereas Red Hat is not. ”
wrong. both are offering workstations
“4. Solaris is significantly cheaper then RH Enterprise. ”
depends on tco and roi
“5. Sun already has case studies of customers who they have migrated away from Linux, and the customers are much happier with Solaris than they were with Linux.”
solaris to linux migration is much more common. case studies exist for the same all over the net.
>>5. Sun already has case studies of customers who they have migrated away from Linux, and the customers are much happier with Solaris than they were with Linux.
Like Oracle who are in the process of switching over 14,000 workstations from Solaris to Linux citing price and performance as the main motivators.
“Oh, almost forgot. RHEL kernels already contain backports of a whole lot of scalability enhancements from 2.6.”
Not most of them at all. The whole VM/io subsystems, for example…
More interesting paragraphs:
“On static and dynamic webbench, Solaris 10 is at par with RHEL AS3 on 2 CPU v20z” <- So, what would happen if we put linux 2.6 on the scene?
“tell me after they do it. they have been telling that for quite sometime. what about sco?”
Remember that Solaris is based on BSD. And BSD has immunity from lawsuits because of the agreement between UCB and Novell.
Sun’s lawyers have already stated that SCO cannot stop Sun from open sourcing Solaris.
“they arent locked into one vendor with linux either”
Technically they aren’t. But for practical purposes they are. There are enough subtle incompatibilities between different Linux distros that most companies will only support Red Hat, and maybe SuSE as well. Since most Oracle customers for example, need support from Oracle, they have to run one of the supported Linux distros. Which typically means they have to run Red Hat.
“solaris to linux migration is much more common. case studies exist for the same all over the net.”
True. But this was before Sun started to wake up and smell the coffee and realize that they had to make a concerted effort to put Solaris up against Linux. It will be interesting to see what happens with this change over the next year or so.
“Like Oracle who are in the process of switching over 14,000 workstations from Solaris to Linux citing price and performance as the main motivators.
”
Like the Chinese government, for which Sun is filling orders for two million JDS systems (which likely will be Solaris based since Sun seems to be backing off the Linux version of JDS a bit)
“Like Oracle who are in the process of switching over 14,000 workstations from Solaris to Linux citing price and performance as the main motivators.
”
By the way. This was also before the Solaris pricing shakeup.
Solaris is now cheaper than Red Hat. And Sun is convinced they are going to get Oracle back.
>>Like the Chinese government, for which Sun is filling orders for two million JDS systems (which likely will be Solaris based since Sun seems to be backing off the Linux version of JDS a bit)
Sun said 200 million computers and that they would run LINUX!!!
http://sfgate.com/cgi-bin/article.cgi?f=/c/a/2003/11/18/BUGRM34CUC1…
Still nothing. Just more BS from Sun.
>>By the way. This was also before the Solaris pricing shakeup.
Whats your point? Sun can’t compete on technical merits with Red Hat?
The solutions were evaluated and RH’s offerings were found to be superior.
Remember that Solaris is based on BSD.
—-
Please talk about what you know. sunos was based on bsd. solaris is sysv based and has code licensed from sco.
“Since most Oracle customers for example, need support from Oracle, they have to run one of the supported Linux distros. Which typically means they have to run Red Hat. ”
not really. suse is a clear alternative with a strong support channel thru novell now. userlinux might be a viable choice is bruce perens get his act together
” It will be interesting to see what happens with this change over the next year or so.”
this I will accept. I am closely watching what Sun is doing. I sure hope that Jonathan doesnt used his blogging for abusing HP, Redhat and IBM like this
Like the Chinese government, for which Sun is filling orders for two million JDS systems (which likely will be Solaris based since Sun seems to be backing off the Linux version of JDS a bit)
_—–
Wrong there. Chinese goverment has really signed two million systems and they are biased towards Linux. they are moving into that for all govt operations with their home made red flag distro. they arent taking in solaris. I know this for sure
And Sun is convinced they are going to get Oracle back.
—
problem is not whether sun is convinced or not. question is whether oracle is and oracle did their conversion based on technical merits and not system cost. they arent going to do the migration back yet again even if it didnt
Hi,
Not to bash RH, but I think you’ll find that Simba has a point about being locked in – sure, you might not need to sell your soul and pledge undying allegiance to the Altar of RH, but if you purchase their enterprise products, you’ll find there is a tendency to be stuck with them. Of course, if you like RH products, there’s nothing wrong with that. (And yes, I know about WhiteHat, Tao and CentOS – not the same – for Rh, it’s all about support/services, and making the higher-ups happy).
Also, it’s a rather annoying trend that many companies are now saying they only support RH/Suse etc. (IMHO, Suse seems a little less enterprise-oriented).
spank_da_monkey: Thanks for the links. The first one was kinda interesting, particulalry the article discussion pages. In the static Apache benchmark 2.4 actually beat 2.6, whilst 2.6 won out by 5.5 percent with dyanmic content
According to some guy on the discussion page, the improvements in dbench can probably be attributed to the improved ide i/o, which means it might not be such a huge gap with scsi drives (of course, it should still be better).
Anyway, good to see this being discussed – I actually submitted this article to Eugenia a few hours ago to have you guys pick it apart and analyse it, although I think Sunay may have beat me to the punch *grin* (see his blog).
bye,
victor
<<<Actually, what I find funny is how predictable you Linux zealots are.>>>
Actually, what I find funny is how predictable you Sun shills are.
Sun compared their “next gen” product against a shipping Red Hat product based on a several year old 2.4 kernel. If Sun had the cojones to benchmark against Red Hat or SuSE enterprise beta products using the 2.6 kernel, I might care enough to read the details.
Even if there was some minor temporary performance advantage, Sun’s product is closed and proprietary and Sun is the ultimate “flip-flopper” on x86 support. Most are smart enough not to go down that road, as Sun’s financial failures demonstrate.
This is as bad as it gets – measuring different hardware platforms, running old OS against beta OS with different software set’s on top of each platform to produce meaningless graphs.
I’d expected more from sun . . . They should sack whoever produced this.
Spank_da_monkey: Dude, chill out a bit =). And no, it’s (2 million Chinese systems) not “BS” (can’t you use ‘rubbish’ instead?), it was reported in the news, contract signed, paraded around by Sun (*grin*).
Also, as far as I’m aware, JDS is the *desktop* on top of the OS – whether it has a Linux or Solaris base is irrelevant. From what I understand of SUn’s aims, it should be indisinguishable (or thereabouts).
Also, what do you mean “evaluated…Rh’s found to be superior”. I think you’ll find most people check out the price-tag before considering whether to buy. Sure, Ferrari’s are faster than Supra’s (and look waaaay better =) ), but which one’s better value?
Bye,
Victor
Also, it’s a rather annoying trend that many companies are now saying they only support RH/Suse etc. (IMHO, Suse seems a little less enterprise-oriented).
—
suse is no less enterpise focussed. you got to be kidding me there. i would love to see a debian based distro like userlinux come up but you will have to admit that RH is doing a good job with sticking with open source and you can name none of their specific features as lock in.
>http://www.google.com/search?q=linux+kernel+2.4+vs+2.6+benchmark&am…..
>http://www.2cpu.com/articles/98_1.html
>http://kerneltrap.org/node/view/903
Did you read this btw ? Yes, they show 2.6 to be faster. Much less than I assumed actually (less than 10-20%). So you proved my point. Thanks.
>>Spank_da_monkey: Dude, chill out a bit =)
😛
>>And no, it’s (2 million Chinese systems) not “BS” (can’t you use ‘rubbish’ instead?), it was reported in the news, contract signed, paraded around by Sun (*grin*).
There still has been no deployment and considering how much Sun is trying to bolster JDS (proably a Solaris push), I doubt that this huge influx went unnoticied and wasent immedeatly publicised/flaunted by Sun.
Sun has nothing.
>>Also, as far as I’m aware, JDS is the *desktop* on top of the OS – whether it has a Linux or Solaris base is irrelevant. From what I understand of SUn’s aims, it should be indisinguishable (or thereabouts).
The question then has to be asked, why hasent JDS2(linux-based) been updated? Why is it so horribly out of date (linux-kernel 2.4.19). Linux is not their priority, they are a Solaris whore now.
>>Also, what do you mean “evaluated…Rh’s found to be superior”. I think you’ll find most people check out the price-tag before considering whether to buy.
“Evaluated” encompasses migration costs/learning curve/productivty gain or loss, etc…..
and yep Red Hat won!<—— Gave Sun a royal spanking!
Nope just posted the first two links. I thought you were saying you couldnt find any benchmarks
. Heres an intresting one.
“…the Apache server with 2.6.0-test5 kernel more effectively used system resources and served six times more Web pages than the 2.4.18 kernel did.”
http://www-106.ibm.com/developerworks/linux/library/l-web26/
>Nope just posted the first two links. I thought you were saying you couldnt > find any benchmarks
. Heres an intresting one.
> “…the Apache server with 2.6.0-test5 kernel more effectively used system resources and served six times more Web pages than the 2.4.18 kernel did.”
They were testing apache on an 8 way system, which means that 2.4 kernel would suck so bad on it that it wouldn’t even be funny — Solaris would absolutely smoke Linux on an 8+ way system even with 2.6 kernel. Remember Solaris has multithreaded kernel and has a pretty good advantage over Linux in multiprocessor space. The only place were Solaris had a weakness was single/dual processor performance and it’s all history now — Solaris is better across the board
Buddy I don’t have a copy of Solaris 10, so your comments about Solaris smoking Linux are pretty much hersay…to me. Anyways my point is Sun’s benchmarks are in my mind and in many otheres, void. It compares a product that isn’t avaliable(Solaris 10) and one that is 3 years old.
spank_da_monkey’s quote of the day:
The benchmarks do not echo market reality, its Solaris 9 vs RHEL 3 out there.
> It compares a product that isn’t avaliable(Solaris 10) and one that is 3 years old
Errr, apparently you don’t know much about the product you’re trying to defend. The official release date of RHEL 3 is 2003/10/22, which is less than one year old. And on the subject of Solaris 10 being unavailalbe, there are already production level deployments of Solaris 10 as a part of beta program. If you want to try Solaris 10 yourself, just pull it down from Early Access at http://wwws.sun.com/solaris/solaris-express/sol_index.html, which by the way something you can’t do with RHEL (RH tries to keep RHEL away from public).
>>Errr, apparently you don’t know much about the product you’re trying to defend. The official release date of RHEL 3 is 2003/10/22, which is less than one year old.
LMAO what am I smoking? You got me there buddy, I heard 2-3yrs from a post in this thread.
>>And on the subject of Solaris 10 being unavailalbe, there are already production level deployments of Solaris 10 as a part of beta program.
Just like RHEL 3.94 “nahant”. So why didn’t Sun benchmark RHEL 3.94?
>>which by the way something you can’t do with RHEL (RH tries to keep RHEL away from public).
http://rawhide.redhat.com/pub/redhat/linux/beta/nahant/iso/
Red Hat has always been very open to their development process.
Reading the benchmark, analysing it and I came to the conclusion: Sun’s benchmark is pure marketing since it is based from beta version vs final version. Solaris OS will available next, so are the incoming RHEL4 which is based on kernel 2.6 and also the incoming Fedora Core 3.
If Sun were honest, they should compare Solaris 10 vs RHEL4. Remember the stupid comparison from Microsoft with Windows running on desptok vs Linux(which distro?) running mainframe.
I’ve been running it on my desktop. It’s noticably faster than Sol9, and seems zipper than the FC3 beta I was running. Linux’s disk performance is a bit better, but otherwise, things feel snappier on Solaris now than they do under Linux. I’m very impressed so far.
Why does it have to take almost 1.5 hours to install on an XP2400, 42X? And then if you are not lucky, you’ll have to track down on a differnt machine, of course) a 3rd party network driver in order to jump on the net, but only if you get to mount the floppy, which seems to differ from version to version.
I was really looking to the new incarnation of Solaris so I can install it on a 2×2.8 Xeon, 2G, especially after reading somewhere (anybody knows the links) a significant performance increase over RH. I even listened to Sun’s online presentation (a few weeks ago). But the 1.5 hour wait gave me enough time to realize that in spite of all Sun’s efforts (especially after pulling the rug temporarily from under x86 users), Solaris for x86 is not meant to be deployed in a real environment and only to be used as a teaching/learning tool (resume filler) before jumping to Solaris on Sun hardware.
Maybe Sun should take a look at RH’s (or any other distro) installation process. Having to wait 1.5 hours and in the end not having a networkable system is not acceptable in the real word.
Something tells me you’re having some driver issues during install with both your CDROM and the network card — chances are the install is not going at 42x your CDROM is claiming to support (most likey working in some legacy mode). I have installed Solaris 10 numerous times and on the well supported hardware (check HCL) the install times are not much different from any Linux install with the same number of install cd’s. I agree that the support for a good number of PC NIC’s out there need to be improved on Solaris, but supposedly it is coming as Sun is working with IHV’s to get the drivers ported to Solaris x86.
I haven’t had problems with long installs, either. I haven’t clock an install from start to finish but I venture its about 45 minutes.
And solaris on x86 is really fast and rock solid, especially Solaris 10.
just posting dumb benchmark results is no good;
they gotta post how they go them!! and make sure some one who suspects the way they did can reproduce the results.
Now if they are actually comparing comparable stuff at all is an entirely different issue.
🙂 benchmarks published by any company with the sole purpose of projecting their product in a better light
) thats called an advertizement not a benchmark comparison.
any one ever notice the microsoft ad that flashes in some OSnews articles? it claims that windows server2003 is several times better than linux for a particular task. why dont some one verify that and comment?
this is no different than that, except maybe sopme agency is paying OSnews for placing that ad in the articles.
~ram
if sun suddenly ‘makes up’ these benchmarks then ***you can legally sue them*** for a lot of money…
so why would sun do this?
oh wait, your spreading FUD against solaris arent u??
how lame
Heh, 2.6 outperforms 2.4 but i dare you to provide details on how 2.6 outperforms solaris.. how lame.. with your fud.. ewwww
honestly 2.6 isn’t even as good as people said it woudl be and solaris still outperforms it so you really have no basis for saying crap.
Solaris is not like Linux when it comes to hardware, if it is not on the HCL, it more than likely WON’T work. I use Solaris Intel at home and I usually get install times around 30 minutes (dual 1 GHz PIII, dual 40 GB WD Special Edition drives , and a 52x CD ROM). This is on a SuperMicro P3 motherboard with a VIA chipset.
If your disk performance is bad, are you tuning the disks? If you are using IDE, changing the blocking_factor in /platform/i86pc/kernel/drv/ata.conf from 0x1 to 0x10, save and reboot. With Solaris 10 you should not have to change anything for DMA transfers, but I would check /boot/solaris/bootenv.rc and ensure the setprop ata-dma-enabled line is set to “1”. And check /var/adm/messages for any IDE errors.
If you are using SCSI disks, are you using maxphys (/etc/system) and sd_max_xfer_size (sd.conf)? See the following for more information:
http://www.samag.com/documents/s=7610/sam0210j/0210j.htm
http://www.solarisinternals.com/si/index.php
you obviously don’t even look at the news do you? The open sourcing thing was recently and it takes awhile for things to go thru with huge corporations.
they already made statements on SCO and SCO backed off of the statements that they made about sun, it wasn’t even a big exec from SCO that said that either. Sun welcomes a lawsuit and knows SCO will fail ..
what carp, linux isn’t all that good in arch. design… HURD is probably the best designed OS. Solaris was created from SysV and some parts of BSD which are designed to run on the server.
Linus created linux just as something better than minix never did he intend to use it as a big time server OS. It was complete accident that linux even got this popular, linus had no idea there’ quotes from him about it. linux started to get big after companies invested millions to make the system even somewhat compete with the commercial UNIX distros..
IF you were an advanced unix administrator you would know how the kernels are designed and know that linux has major faults… with that being said freebsd has bigger faults but they are fixing it just like how linux is fixing theirs..
solaris is pretty much the most scalable open OS in the world i believe, scales much better than linux, this has been proven and any sys admin would be shitting on themselves if they say linux is surpior of them all.
with that being said, linux is evolving FAST, so fast that’s being adopted all other the world.. tell me, what would you do? stay on an OS that was closed at the time and wasn’t growing as fast as linux, possibily loosing ground? or would you switch to an OPEN linux system that is technically improving VERY FAST and could possibily surpass solaris and the other big OS’s FAST? Oracle made a smart decision… however… they probably made that decision before sun said they would open source solaris.. so it appears that sun is doign what it can to make sure solaris stays in front of linux and thank god they are opening it.
now look over what i just typed if you are ready to type some evil, nasty comment to me and ask yourself, can i prove that he was wrong?
Woah, I think we’re taking this conspiracy thing a bit too far. Ddd’s right – you can *not* intentionally lie and mislead the public in a press release like this, particularly if you are a large, publicly listed corporation like Sun (or MS for that matter).
I haven’t checked it out yet myself (why would I bother, it’s MS *grin*) , but I think if you examine that MS advertisement-benchmark yourself carefully, you’ll find somewhere in the fine-print the exact testing conditions they used, or at the very least an email address where you can request additional information. So they might well be right when they say Windows outperforms Linux under those conditions. Of course, we all know Linux/Apache completely flounces MS in real life conditions =).
With Sun, I don’t really see much obviously wrong with their configuration. Of course, my experience/knowledge on this thing is limited, so I’ll leave it to other people to debate this. However, I have a suspicion several people haven’t bothered checking out the “Benchmark Configuration Data” page at http://www.sun.com/2004-1012/feature/bench.html .
And no, RHEL 3 is NOT comparable to Sol9, which I believe several years old. And as somebody else already pointed out, RHEL is less than a year old, whilst Sol10 is already deployed in several companies (apparently they subscribed to some Sun Trial Program), and also freely available for download. My experience so far is that it’s quite snappy and responsive, as well as nicely polished.
bye,
Victor
They’re outperforming redhat AS 3.0 by 15-35% in the 2-way field…ok. Those would be nice numbers, until you realize that redhat as 3.0 is based in the 2,4 kernel. 2.6 already outperforms 2.4 by a big margin, I’d have expected more from solaris….
Pulease! the only features of Linux that are noticable from 2.6 is when you start scaling beyond 8 way, until then, you’d be lucky to see a damn difference between 2.4 and 2.6. The fact remains, RHEL 3.0 includes ALOT of 2.6 features already, NTPL sitting at the top of that list.
Suck in your bottom lip and accept it; Solaris IS improving on x86, and whether you like it or not, it IS becoming a damn good competitor for not only the commercial linux world but Windows as well.
Some of Linux’s not so impressive performance numbers on Opteron are likely due to GCC and not Linux itself.
GCC is not known for producing very efficient code on 64 bit systems. Sun is likely using a propreitary compiler to build Solaris 10 that is probably much more efficient at producing 64 bit code.
Come on, the crappy quality of code from GCC has more to do with the complexities of writing a good compiler for RISC rather than an issue of Opteron being 64bit.
RISC is like diet-EPIC, EPIC you could describe as an extreme version of RISC, a really-really-really-really reduced instruction chip where virtually EVERYTHING is dumped onto the compiler. A good RISC compiler is already tough to write already, EPIC is even harder.
Conclusion: The performance difference has NOTHING to do with GCC and everything to do with a superior fine grained system designed from the ground up to be built-like-a-brick-shithouse.
solaris to linux migration is much more common. case studies exist for the same all over the net.
Correction: Move from crusty 15 year old SUN Microsystems servers to Linux is common over then net.
what carp, linux isn’t all that good in arch. design… HURD is probably the best designed OS. Solaris was created from SysV and some parts of BSD which are designed to run on the server.
Regarding design, I would say Mach/OSF was probably the best designed, which makes up the base of Tru64 UNIX.
Regarding Solaris, just a minor correction; SunOS was based on BSD where as Solaris is SunOS + SYSV which SUN licensed off the old UNIX Labs when it was part of Novell.
IIRC, they paid around $30million or so, which was a pretty sum back in the late 1980s, which gave them open slather in what they were able to do with it.
I wouldn’t use Solaris if they paid me to use it.
How about apache 2.0 ? why sun use apache 1.3?
redhat come with 2.0 and not with 1.3!
Actual benchmarks are much better than the ‘Redhat Sucks I hate RedHat RedHat gets my panties in a big bunch RedHats mother was a whore’ approach to marketing than Mr J. Schwartz seems to favour.
If Solaris outperforms Linux, then Sun should show it off.
They have clearly done a lot of work to improve their standing in the low end of the server market and don’t deserve to be bashed for honest work with measurable performance results.
> How about apache 2.0 ? why sun use apache 1.3?
> redhat come with 2.0 and not with 1.3!
Linux kernel is even less likely to do well with apache 2.0 comapred to Solaris. Actually Linux has an advantage over Solaris on apache 1.3, since 1.3 version is not multithreaded and relies heavily on forking. I assume we all know that fork implementation on Linux is quite a bit more lightweigt than on Solaris giving Linux an advantage on any process that is relying heavily on fork/exec. Multi-threading implementation on Solaris on the other hand is very well performant and has proved itself over the years. Bottom line, Solaris is likely to perform even better with Apache 2.0 and Linux is likely to perform even worse on Apache 2.0 compared to 1.3.
These are a small set of unverifiable graphs with little information about the setups. What’s more it is performed by Sun.
If anyone believes these benchmarks, then they’re already sold Sun supporters who don’t need any more encouragement. I don’t see the point.
Actually solaris 10 ‘early acces’ version has been available for a little while now, you can get it for free and test it.
I have the 3 install discs lying around for almost 2 weeks now but havent installed it yet…
Solaris is BSD based…!!??????
Come on you really not know what you are talking about
you are jst a troll. Defending itself with unproven and untrue infromation
Shame on you.
Solaris outperform Linux, I don’t see the news here… if it’d be the other way around I’d be surprised. Solaris has always been the better… it’s simply a question of money. Sure if you use Gentoo or whatever maybe it’s cheaper. But buying from Red Hat is just plain stupid if you aim for performance…
Solaris is BSD based…!!??????
I don’t know if that statement is so off the mark. Bill Joy, co-founder of Sun, helped develop BSD.
1993: Sun announced that SunOS, release 4.1.4, would be its last release of an operating system based on BSD. Sun saw the writing on the wall and moved to System V, release 4, which they named Solaris. System V, release 4 (SRV4) was a merger of System V and BSD, incorporating the important features found in SunOS.
unixed.com/Resources/history_of_solaris.pdf
there’s too many variables in these benchmarks to make them very interesting, the hardware difference being the biggest one. I’d like a comparison where the only difference is the kernels being used, but I guess they are mainly trying to prove Sun solutions vs. Red Hat solutions.
I’m afraid all we’ve got here are a bunch of graphs that are about as meaningful as those TCO ones you see that have Windows and Linux on them, and surprise, the Windows one is lower. If Solaris 10 really is that much better than Linux then we would have seen it in all its glory. The blog is desperately trying to garner some interest in this, so we’ll have to see what happens. The question is, unless Solaris is an absolute country mile ahead of Linux down the road in terms of performance, are people actually bothered by silly bar graphs?
Given that RHEL 3.0 is kernel 2.4 based (albeit with some backported updates), and Sun have had the ability to dictate the exact hardware, tuning and what they actually measure and the circumstances, I’m not impressed.
After briefly switching to Linux, the company returned to the Solaris OS, lured by the Solaris 10 OS’s astounding performance. The ISP now sends 25,000 messages per second on the same AMD Opteron processor-based hardware and has abandoned Linux.
No one abandons a platform over a bar graph and no one switches twice. Given that Solaris 10 isn’t even released this ISP must have been paid to do a test deployment of Solaris 10. It doesn’t say how many messages per second (whatever the hell that is) the Linux servers were doing.
This is tit for tat, and is not a war Sun is going to win.
“As much as 58 percent lower”: Red Hat Enterprise Linux AS, including 5 x 12 support, is $1,499 per year. Solaris 9 OS two CPU RTU is $250. 24 x 7 support is $540 per year.
Well, the Red Hat pricing page is far easier to understand and they know who they’re selling to. If you’re doing some simple deployments it is very easy to understand, and if you have hundreds of servers you call them.
The Sun page, on the other hand is a mess. Why would I be interested in Solaris 7, 8, or 9? Solaris 10 isn’t even there when RHEL is being sold! The $250 is the absolute basic x86 installation, and is not enterprise grade. I couldn’t find the 24×7 $540 option. The enterprise support installations are all for 100 licenses or above and nobody in their right mind simply orders them off a web page at face price. You’ve also got documentation sets sold separately, encryption packs(!) sold separately and God knows what else in the license sold separately. There’s also a load of binary license restrictions I couldn’t be bothered to read. Since Red Hat’s software is GPLd and I’m paying for the support, that’s all I need to know.
Having looked at Sun’s prices in more detail than I have done before, I can definitely say that Sun is most certainly not cheaper than Red Hat in the comparisons they’re making. Sun are trying desperately to make distinctions between what they are trying to sell as enterprise grade and the basic commodity two-way servers they sell. Red Hat are bringing the enterprise stuff to the low-end (you can even run RHEL on Power for that price) and selling it for marginally more. Red Hat still have growing pains in all of this, but they’re heading in the right direction. The Sun stuff just doesn’t stack up.
Sun need to work out what their target market is and what those customers expect. Sun are going to try and keep their large-scale enterprise deployments price tags whilst giving the appearance of being cheap with basic x86 servers that will be treated like they’ve fallen of the back of a lorry by the Sun sales and support staff (even the Solaris based ones). They did this with Cobalt and they’re going to fail again.
http://store.sun.com/CMTemplate/CEServlet?process=SunStore&cmdViewP…
Cluster :
http://store.sun.com/CMTemplate/CEServlet?process=SunStore&cmdViewP…
1 Server License for Netra t112x, Software & Documentation [+ $2,995.00 each]
Firewall :
http://store.sun.com/CMTemplate/CEServlet?process=SunStore&cmdViewP…
Software & Documentation [+ $14,995.00 each]
Development tools :
http://store.sun.com/CMTemplate/CEServlet?process=SunStore&cmdViewP…
10 RTU Slim Kit [+ $25,450.00 each]
RHEL 3 AS come with all of this you have the _source_ and no per user licence (Development tools) !
Anyone have a .torrent?
We’ll see how it compares against Gentoo 64-bit on an Athlon64 2.2Ghz laptop.
good question anonymous, i went to Sun’s site only once to see about downloading Solaris when they first offered it for X86, and they wanted me to register, sure i could have done so with bogus info, but why bother – Linux is available anonymously no regestration necessary, a couple of years ago when XP was about to be released i read about the product activation and it made me realise that i am just jumping thru hoops for a stupid OS so i made the switch to Linux and loved it, no need to register, no product ket, no activation, and since i made the Linux switch i havent looked back.
i was going to give Solaris a try but seen that registration page and just closed it in disgust, i refuse to tolerate such nonsence…
i dont mind paying a few bucks for a OS, but they do not need all that persojnal info, what are they going to do with it? send me spam later??? i finally got my email cleared of spam and am not going to open it up for that again…
Considering the 64-bit code has not been released to the public yet, it might not compare too well. Build 63 is 32-bit and has none of the AMD optimizations (to my knowledge).
One of the interesting things about Solaris though, is that it has not billed itself as UNIX System V since Solaris 8. Solaris 7 was the last release of Solaris to do this.
So it seems that Sun is trying to distance themselves from anything that is officially UNIX System V. This makes be believe that Sun has removed any Sys V source code and that Sun actually can make good on their claim to open source Solaris 10.
“How about apache 2.0 ? why sun use apache 1.3?
redhat come with 2.0 and not with 1.3!”
It can be argued that Apache 1.3 is more stable than Apache 2.0. As another poster pointed out, Apache 1.3 is not multithreaded, and relies entirely on forking itself to handle requests. This is why Apache has been known for being extremely stable, but has also been known for being one of the slower Web servers out there. processes are extremely expensive compared to threads.
Apache 2.0 will be faster then Apache 1.3 because it uses threads. However, this has a drawback. Because threads share memory space and such, a single misbehaving thread can bring down the entire Web server. With processes, this is much less likely to happen even if one process misbehaves since processes each run in their own memory space.
“this is crap just like your claim that solaris is based on bsd. threads are much more safer to crashes than processes and linux treats very well now due to NPTL.”
Heh… Guess you have never heard of a race condition then huh? Or had to debug a random problem with a variable losing its value because some bad pointer arithmetic in a thread overwrites the end of an array at random times…
Clearly you have virtually no understanding of threads.
BTW, I am not talking about crashes because of OS problems. I am talking about things like a synchronized thread failing to release a lock and thus blocking other threads in the same application from running. (Due to the a programming bug and such.)
But if you knew anything about threads, you would have known that.
Let me give an example of what I mean.
Suppose I were to find a hole in Apache that allowed me to lock it into an infinite loop remotely.
Under the process forking model, the only think I can do is lock the one process that is serving my request. This won’t affect the rest of the Web server because Linux can pre-empt that process to run a different once.
Now suppose I were able to find a hole that allowed me to lock a synchronized thread in a multi-threaded Web server into an infinite loop. Now I can bring down the entire Web server. Because until that thread finishes what it is doing, no other threads in the Web server can run. And Linux cannot pre-empt that. If it could, it would defeat the whole purpose of the synchronized thread. The synchronized thread is synchronized because it has to finish what it is doing before any other thread can run in order to prevent race conditions and such.
Linux itself would continue to funtion normally, and so would any other services such as email and such. But whe Web Server would be brought to a halt because it can’t run any other threads until the synchronized one finishes what it is doing (which will never happen since it is locked in an infinite loop).
Get it now?
Now suppose I were able to find a hole that allowed me to lock a synchronized thread in a multi-threaded Web server into an infinite loop. Now I can bring down the entire Web server.
—-
this is wrong.
1) apache can use both the prefork model and the other one. your claim that somehow apache 2.0 is problematic because it uses threads by default is false because this behavior is configurable
2)) NPTL allows treating of threads as processes which are preemptible and wont allow hogging resources. read the redhat people whitepaper on this if you want to know me.
obviously you dont know the specifics while claiming stuff based on your general understanding of threads which i suspect are based on java threads model on 1.4
“apache can use both the prefork model and the other one. your claim that somehow apache 2.0 is problematic because it uses threads by default is false because this behavior is configurable”
I should have clarified and said when it is using the threading model.
“2)) NPTL allows treating of threads as processes which are preemptible and wont allow hogging resources. read the redhat people whitepaper on this if you want to know me.”
Only non-synchronized threads can be pre-empted. And that is nothing new. Synchornized threads cannot be pre-empted. To do so can be disastorous. You can model this with the classic consumer / producer thread model. What if the consumer tries to read data from the producer before the producer has actually produced that data? It’s going to a bad value because it will read the old data. So the producer has to finish producing the new data value before the consumer can read it.
You can model this with the classic consumer / producer thread model. What if the consumer tries to read data from the producer before the producer has actually produced that data? It’s going to a bad value because it will read the old data. So the producer has to finish producing the new data value before the consumer can read it.
—
this can be avoided in a fool proof manner. please try hogging apache by just using threads rather than processes. come back and tell me threads model was easier to crash
“this can be avoided in a fool proof manner.”
It can be avoided by using polling. But polling is inefficient and is how we used to do things like this before we had threads.
“please try hogging apache by just using threads rather than processes. come back and tell me threads model was easier to crash”
As I said, to do this I would have to find a bug that would allow me to lock a synchronized thread into an infinite loop. I’m not saying there are any known bugs that allow this. And I really don’t have the time to sit around and try to look for unknown ones.
It can be avoided by using polling. But polling is inefficient and is how we used to do things like this before we had threads.
—
there are several other methods. see recent usenix papers for details
“. And I really don’t have the time to sit around and try to look for unknown ones. ”
unsubstianted claims are useless
“unsubstianted claims are useless”
Most ideas about security and safety in software are based on theory. Not on actual tested holes. So theoretical possibilities for security problems are not only useful, but essential to good design.
Actualy, apache 2.0 (on unices) uses by default a hybrid model.
It uses a few processes with plenty of treads.
So if one tread goes mad, it won’t brind down the entire webserver.
theoretical possibilities for security problems are not only useful, but essential to good design.
—-
not when it totally doesnt apply in practise. your theorotical problem with threads in inapplicable to apache 2.0. this has already been discussed in their mailing list. everyone agrees the new model is good. even rasmus of php who gets poor performance agrees this is a better model.
May be I’m alone but I think this benchmark is good.
Ok, this is pure marketing, RH3.0 which have been out for some time is compared to Solaris 10 which is to be released.
So Linux Kernel is old, Apache is still 1.3 and a benchmark from one of the competitors is not something I would rely to make a buying choice.
Anyway, this is good because UNIX market is becoming hot again !
When was the last time Sun felt obliged to improve as much its OS than between v9 and v10 ?
They are having competition and I think this is making them improve.
Also, I think it is good to have them compete on the performance and features field, because this field is far more sane than the IP/patents one.
In the end, I guess Linux will dominate because this is a far easier OS to get running on any kind of trashy PC hardware, it is a very common first UNIX for learners, and lots of hype applications are hard to build (or run) on non-Linux Unices (which is bad of course). Solaris will probably be confined to a more standardised environnement, for people buying some “reliability insurance”.
I think we should also appreciate the fact that Linux has some competition and I would not like to work in a “Linux, Windows or nothing” computer world.
For those complaining of unfair benchmark, Linux and Solaris 10 are free to download, so run your own !
BSDs, Unices, and other OSes everybody is welcome.
Isn’t it what OS news is about ?
cheapskate: Dude, what the? They’re not asking you for any money, they’re just asking for your name and email address. And no, they won’t send you spam. The only emails I ges from Sun are when I suscribed to the their NC ’04 Webcast, when I specifically asked (alright, fine, they suggested and I said yes) to be reminded one week before the webcast. I get more spam for VMWare asking to register for silly VMWorld event that I have to pay $80 for…and I don’t see anybody whining about download the VMWare trial
http://store.sun.com/CMTemplate/CEServlet?process=SunStore&cmdV…
Cluster :
http://store.sun.com/CMTemplate/CEServlet?process=SunStore&cmdV…
1 Server License for Netra t112x, Software & Documentation [+ $2,995.00 each]
Firewall :
http://store.sun.com/CMTemplate/CEServlet?process=SunStore&cmdV…
Software & Documentation [+ $14,995.00 each]
Development tools :
http://store.sun.com/CMTemplate/CEServlet?process=SunStore&cmdV…
10 RTU Slim Kit [+ $25,450.00 each]
RHEL 3 AS come with all of this you have the _source_ and no per user licence (Development tools) !
1) Why are you comparing GCC with a commercial compiler? GCC is available on Solaris, next time you compare, actually be honest in your comparision; the way you compared that pricing was like Microsoft comparing its packages with Linux.
2) You don’t pay CAL (Client Access Linceses) for Solaris.
3) If you are going to deploy a Solaris based solution, you would have the nouse to use the JES package which includes everything one needs to get up and running; available for both x86 and SPARC servers.
4) Who gives a shit if you have the source; no one outside the programming community actually give a shit if they have the source; 9/10 times, the customer will ring up Red Hat regardless of the fact that they have the source code. Administrators administrate, they don’t sit down for hours on end sifting through a coredump finding how the application failed then sift through a few thousand lines of code to fix the problem.
Read and repeat, computers work FOR customers, NOT the other way around. Computers WORK to IMPROVE efficiency within an organisation, the core focus of an organisation ISN’T IT, it is selling products and maintaining customer relationships; something that alot of IT geeks here just don’t get.
Hello,
Had fun reading this, thanks. Just a few more points.
Sun clarifies some of your questions in his blog, worth a
read at:
http://blogs.sun.com/roller/comments/sunay/Weblog/solaris_vs_red_ha…
I found the criticism of the benchmarks quite interesting.
When Schwartz claims a performance advanatage over Redhat
he is dismissed because of insubstantiation. When some
substantation is provided it is dismissed because of bias.
Well, yeah, I am sure there is some bias but the facts are
not allowed to be altered (as some posters have already
stated). Also, the complaints about comparing an OS which
is not even in Beta to a production OS. Solaris 10 is at
the DISADVANTAGE here is as it isn’t finished yet. I can
only imagine what would have happened if Sun had posted
benchmarks of Sol9 beating a beta Redhat release.
The facts are that Sun are competitive on x86 hardware in
terms of price AND performance with commercial offerings of
Linux which have the necessary ISV certification to satisfy the majority of customers (i.e. Redhat).
This was was the point of the exec statements recently in
the press and in blogs ( not an attack on linux as most
people without a clue depicted it.)
I couldn’t agree more with kaiwais last post.
First, I do appreciate that Sun is trying to move the discussion towards facts (however biased) and away from Schwartz’ FUD. You will _never_ win hearts and minds back with his approach.
<<<The facts are that Sun are competitive on x86 hardware>>>
The facts are also that even if there is some small temporary performance advantage on very rare hardware, Sun’s product is closed and proprietary and Sun has abandoned low-margin Solaris profits from x86 support to protect high-margin SPARC hardware profits several times in the past.
Sorry guys “IT geeks” _do_ get it, and increasingly what we aren’t getting is Sun stuff.
The benchmark is not the important part of this Feature Story. The fact that SUN is now competing against their true competitor(s) is exciting. I feel its about time they shifted focus they have obviously screwed up competing with IBM and HP. In other words they are trying to corner the market on mid-range Unix computing where traditionally they have done well in the past. This is not a bad strategy i wish SUN well with this.
I really don’t get why Sun are only gunning for Red Hat here, isn’t there another server OS that runs on commodity hardware they should be competing against? I think it’s called something like “Windows”.
Windows was never a threat to SUN market share. This fable was perpetuated by the heads of SUN for so long everyone believes it as fact. Do you think Microsoft would invest in a company that would form to be direct competition to it in the future? They may battle a little over mid range database servers but thats it.
How about apache 2.0 ? why sun use apache 1.3?
redhat come with 2.0 and not with 1.3!
Because Apache 2.0 is slow and crap compared to Apache 1.3. It also is unstable (even as of now). Very little actual deployment happens in Apache 2.0.
I doubt RedHat comes with 2.0 only, or even suggests 2.0 to its customers.
Do a little homework around the web and you’ll see.
I doubt RedHat comes with 2.0 only, or even suggests 2.0 to its customers.
Do a little homework around the web and you’ll see.
—–
i have done the homework. does millions of sites running apache 2.x count ?. how about the fact that redhat is shipping apache 2.x
The facts are also that even if there is some small temporary performance advantage on very rare hardware,
What are you smoking? The server that were benchmarked in the article (V20z) is a 2-way Opteron server. Nothing about that machine or its configuration is “very rare”. These are as close to commodity as you can get without buying the boxes direct from Taiwan. You’re also missing the point about performance. All performance numbers are temporary. It’s a never-ending process. Processors get faster, code gets re-written, these things are always in flux. The important point here is that Sun considers it a priority to remain price/performance competitive on x86 hardware. Given that this is something people have been blasting Sun for not doing in the past, I’m surprised that you would find this objectionable.
Sun’s product is closed and proprietary and Sun has abandoned low-margin Solaris profits from x86 support to protect high-margin SPARC hardware profits several times in the past.
Sun is Open-Sourcing Solaris, so this is about to change.
Your argument that previous Sun actions, with respect to Solaris, are somehow indicative of future actions is specious. Sun is making a *huge* push to get into the x86 space and there’s no way they’re going to bail out after sinking this much time and resources into the project. I don’t expect that this response will convince you; however, I think it’s important that you realize how rediculous your argument is. It is akin to saying that you don’t trust IBM’s x86 support of Linux because in the past, they ceeded the PC OS market to Microsoft. Pretty silly, eh?
(The following is my personal opinion and may not necessarily reflect the views of my employer)
Why are you comparing GCC with a commercial compiler?
Because GCC is good enough. Many people actually use GCC on Solaris because all the decent software is designed to be built for it.
You don’t pay CAL (Client Access Linceses) for Solaris.
Wow, really?
If you are going to deploy a Solaris based solution, you would have the nouse to use the JES package which includes everything one needs to get up and running;
And this is related to the above how?
Who gives a shit if you have the source; no one outside the programming community actually give a shit if they have the source;
Like the other Microsoft and Sun idiots who don’t understand the open source process, it is the process of opening the source and having people do things with it that is critical. Having people do this, even a few, creates a snowball effect of improvement that indirectly helps everybody – and the person who doesn’t look at the source code. Even if Sun open sources Solaris, if the license they use is too restrictive they will never achieve it.
This is not sun attacking Linux. It is sun bringing competiton to the table. For the most part, does it even matter to the people that don’t even use Solaris. If this benchmark result was entirely true, would you switch? If your using a specific operating system to hanlde a certain task, then this should not bother you at all. If the OS is handling the task the way you want it, then who cares what the benchmarking test shows.
Lets get real Sun needs linux as much as linux needs Sun. Without competition, it stifles innovation, new feature sets, etc… So if sun is saying there OS is better then linux. Next time around the linux developers are going to be saying, “You might win some, but you just lost one!”.
It’s okay if these results are true. This just passes the bucket to
Linux Developers to thread lite and code better.
<<<The server that were benchmarked in the article (V20z) is a 2-way Opteron server.>>>
Percent 64-bit vs. 32-bit, AMD vs. Intel, 2-way vs. 1, etc. Not exactly mainstream.
<<<Your argument that previous Sun actions, with respect to Solaris, are somehow indicative of future actions is specious. <…> I think it’s important that you realize how rediculous your argument is. <…> Pretty silly, eh? >>>
I realize you have a horse in the race (I don’t), but this is lame hand-waving on your part. Sun has a track record of bailing on low-margin x86 when they see it eating high margin SPARC sales. The financial facts haven’t changed since the last time it happened; if anything people should be _more_ concerned about Sun now because when they have punted on x86 commitments in the past, they were MAKING lots of money as opposed to now when they are LOSING lots of money. Not to mention that Sun execs can’t seem to keep a strategy in their heads for more than a month at a time (wear penguin suit/spew Linux FUD, support x86/ditch x86, hate Microsoft/love Microsoft, etc).
If Red Hat or SuSE decide they’re only going to support PPC processors in the future, I’ve got source to the entire stack, the unencumbered right to do anything I want with it, and a wide variety of people and companies to provide support. If Sun sells out (more) to Microsoft and vanishes, I’ll be screwed if I depended on Solaris or Java for anything – which is why we don’t.