Slashdot Mirror
Red Hat Abandons Sparc
Mike Dubreuil writes: "When I went to download Red Hat Linux 7.0 for Sparc I was disappointed to learn RH is dropping support for the sparc processor line. They are claiming that there is a low demand for sparc versions This may be a major blow to the Linux community because Red Hat is the top vendor for commercial copies of Linux. Not to mention that they have the support in place to handle what large companies demand."
Update: 10/02 09:43 PM by CT : Bernhard Rosenkraenz wrote in to say "It is true that we will probably not release Red Hat Linux 7.0 for
sparc. However, this does NOT mean there won't be a Red Hat Linux 7.1 or
7.2 for sparc.
If, at some time, we decide to discontinue
commercial sparc support, we will turn Red Hat Linux for
sparc into a community effort."
Incidentally, /proc is older than linux. I first used it on SVR4, where it had simply the memory image of each process.
Debian still supports Sparc systems, so no big whoop.
"RedHat files anti-trust suit against fellow Linux distributor Debian"
We can't make money off of people when these freaks are giving it all away for free. Not only that, their logo looks better. He felt we had to do *something*.
.... mmm. no...
Lycestra
When I looked at the licensing agreements for the "free" Solaris 7, it was only free as a non-commercial license; if you wanted to make money using Solaris, you had to pay Sun the Big Bucks
--
Solaris is not the ideal OS, it's /proc is primative compared to Linux, it's obviously more expensive. However, the advantages of having the manufacturer support their own OS are obvious. I KNOW all the drivers will work. I know that the X server will have decent performance.
If Linux was mainly hosted on Sparc, it would be different. But if I'm going to use Sparc hardware, I'm going to choose the OS which is most commonly used on that hardware.
First of all, the article is plain wrong.
Second, if we decided to drop sparc support commercially, we'd also try to get a community effort on keeping the port alive going.
Third, this has nothing whatsoever to do with releasing a closed source version. Doing that would do harm to the entire community.
Dropping a rarely used port would hurt a part of the community, but actually benefit others (all the engineering time we're currently putting into fixing up sparc would go to other projects).
This message is provided under the terms outlined at http://www.bero.org/terms.html
Maybe the people tht are using SPARC machines just aren't using redhat... I know that my SPARC isn't and even if it did run RedHat I wouldn't have bought it from redhat.
1. Have you managed solaris? Have you developed for solaris? Have you run solaris? Looks like you have not...
2. The biggest problem of RedHat is QA. Quality assurance. They have tons of bugs all the time. As Sparc is the only BigEndian RedHat platform (alpha is little endian under linux) it will be the first one to bite the bullet if someone in redhat's management looks towards decreasing bug fix tunraround times and improving QA. It is the obvious decision, but it is wrong. Guess why... Because auditing for endianness bugs gives an additional run on general bugs and solves them.
3. And another important factor is that there are very few people that will have both Sun kit and low bandwidth lines. Most of them do not pay for CDs. They do not need to. They can download an entire distro while going for a coffee break.
Baker's Law: Misery no longer loves company. Nowadays it insists on it
http://www.sigsegv.cx/
Uh, I'd question that statement, RedHat have always struggled with support, to date the web support is adequate, if you don't mind waiting a week or so for answers to your questions. Looks like usenet is still the best place to go for Linux support...
I'd just like to point out, we're talking about RedHat. I assume Linux support for Sparc is going to continue, it's simply that there will not be any more official RedHat releases for Sparc.
The university I am at has started using Intel based machines for most things, including running Solaris for the honours students. I beleive this decision was based far more on cost than anything else.
Basically, if people want to run Linux, they'll probably buy Intel. If people are buying Sparc processors, there's a very good chance they wish to run Solaris.
IIRC, Mandrake is based on Red Hat. I bet we'll see Mandrake drop SPARC support in the near future.
Prevent email address forgery. Publish SPF records for y
think back, 6.0 didn't have support for ALPHA or SPARC that came in 6.1
Yes, I know, Debian (being the cool people that they are) has compiled a version that runs on sparc. And in the sense that they support the rest of the Debian distributions, the sparc version IS supported. But who is going to be the primary sort of people that run sparcs? Big companies perhaps?
The fact of the matter is that Debian simply does not have the same kind of real tech-support that RedHat does. You can't call someone at the Debian corporation and demand to know how to get Apache to stop segfaulting. You can't call Debian and ask for a recommendation between web-traffic analysis programs, or credit-card-verification systems. You can't call Debian at all - it's a collective. There is no real responsibility, because no one person is doing anything. There is no headquarters. Instead there are consultants who agree to give Debian advice. Tech support is an IRC channel. This isn't how a big business prefers to do business. There's no culpability, and above all, no one to sue when it all goes south. Debian's form of support is GREAT when they're dealing with individual users - heck, I'd even venture to say it's the best out there. But for a company, they don't want to have to rely on IRC (where someone can just sign off without answering any questions). They don't want to have to rely on consultants with no real tie to the COMPANY that provides the Debian service - and especially not when they charge a minimum of $100/hr.
Exactly. If a company's going to shell out tens of thousand dollars for *each* SPARC server kit, they're not going to balk at paying a few dozen thousands for Solaris, so it will be supported by the original vendor.
Who this hurts, of course, are the individual owners of SPARC kits. Of course, these people who own the kits will probably not be buying their copy of RH at places like CompUSSR....
--
From ultralinux.org...
Here is the link....
Really people... This took 5 seconds to find. Don't make up stuff as you go please.
as well as Alpha, ARM, m68k, PPC, and sparc64, with work also being done on MIPS and PA-RISC. See http://www.debian.org/ports/
http://www.welton.it/davidw/
As long as there is still the other members of LPI6 and the other major distros, ie Mandrake, Linux will still be able to support all that it needs to. Just because Red Hat changes something, does not mean that it will affect the rest of the distros..
öööööööööööööööööööööööööööööööööööö
How Jaded Are You?
It's the nature of the beast.
Got Rhinos?
Sun's Planning (to) Acquire Redhat Computing? Sun's Pretty Angry (at) Redhat's Cancellation? Sun's Petty Architecture; Redhat Cancels? Owwww.
EMUSE.NET
"We're sorry, but the website you're trying to reach has been disconnected."
If you have a SPARC system, how exactly did you get it? Through Sun, right? I thought this was the only way you could get a SPARC machine (correct me if I'm wrong, but as far as I know, there aren't any mom and pop computer stores which build SPARC systems.) This plea for SPARC Linux is almost akin to the anti-Microsoft movement. Dell makes great x86 servers and still bible thumpers complain that they only ship with either Windows 2000 Advanced Server or RedHat; they want it to have BSD or Debian. Sorry to spoil your /usr/lib/bible/thump moment, but that's the way the computer business goes: components (hardware and OSes alike) are provided by the lowest bidder. VIA makes chipsets which crash the operating system; RedHat makes driver modules which foul up the hardware.
"Ancillary does not mean you get to rule the world." --U.S. Circuit Judge Harry Edwards, speaking to the FCC's lawyer
FreeBSD with the SRM firmware was a breaze to install. It was just like the i386 install--easy. No dicking with the ARC firmware was great. Just "boot dka1" (that's my cd-rom drive. Type "show d" at the SRM prompt to get a list of your devices.)
FreeBSD (I've since tried NetBSD and OpenBSD, with similar success on older alpha) seemed to be much more polished on that architecture than other OSen.
Hope this helps.
-Peter
. Penguins Surely Ca
I don't mean to dis the hardware you obviously love--I like it too. I'm also not trying to say you (and the other Alpha Linux gurus) aren't doing a good job. I'm just saying that you don't have the critical mass needed to absorb very many news users at once--which keeps your numbers low in a viscious circle. With more users you get more questions and then more answers. Further new users doing searches find those answers and can jump in easily.
When I search for the answer to a problem with Intel Linux, I eventually find the answer among the hundreds of hits. When I search for the answer to a problem with Alpha Linux, I often don't find the answer among the dozens of hits.
As a datapoint: I've been using Linux for 4-5 years on Intel. I'm very comfortable installing it. I've been using Alpha Linux for 6 months and I've installed it on 3 machines (multiple times on 2 of them). On all 3 machines I struggled for at least a day (and in one case multiple weeks) trying to get Linux installed. Yes, I did eventually get it, with many thanks going to the efforts of the AlphaLinux people (and mailing lists). But it's still nowhere like as well-documented as Intel.
--
Non-meta-modded "Overrated" mods are killing Slashdot
(Hey Ryan! Here's your proof!)
I tried it out last week. Wasn't very impressed. The install finished with neither Backspace or Delete working (All that worked was CTRL-H) and JDK 1.2.2 segfaulted every time I used javac or java. With JDK1.3 for Linux still in the pre-release state, and my primary use is for Java development, I had no choice but to abandon ship and go to Mandrake 7.1...
*If* you really need Sparcs and *if* you can afford them, I don't see why you'd run Linux on them. Solaris seems to do just fine in that case!
Becasue there's a lot of sparc32 hardware out there going cheap (take a look a Ebay sometime), that Sun have dropped support for as of Solaris 8. If you want up to date software (IPv6 support, for example) on your IPX, you have to run something other than Solaris.
By the way, does anyone here has the latest figures from PC-Data?
The truth is, Redhat has always been unable to DELIVER for the sparc platform. I bought 5.1 and 5.2 for sparc in the store. When 6.0 came out, I called to order it, and it was not available and they would not take pre-orders for it. By the time 6.1 came out, 6.0 for sparc never showed up. Even today, no more sparc versions have showed up.
I also asked for a sparc version of their secure web server. I would have bought it if they had it. But they didn't, and even said they had no plans for it. I called again later and ask for it again, and they said no one wanted it when in fact they keep no records of demand at all.
The fact is, Redhat is not interested in supporting other than the one largest platform. As Microsoft is to all operating systems, Redhat is to Linux; they are the "big guys" and as such, they can do, and ignore, whatever they please. The sparc platform isn't in their business plan because their business plan has changed to pursue big corporations (which probably is the right thing to do for their stockholders). They should just tell the truth and say that sparc is just too small a market for their grandiose plans.
now we need to go OSS in diesel cars
there's a lot of sparc32 hardware out there going cheap ... If you want up to date software (IPv6 support, for example) on your IPX, you have to run something other than Solaris.
I hear this kind of reasoning a lot. I just don't get it. An IPX was a kick-ass system for it's day. Now, it's the rough equivalent of a P90 with much poorer support available for video, scanning and many other hardware periperals where under Linux for X86, you have all of the above plus the average P90 is going for even less than that IPX! Why would you waste your time?
The only reason to run Suns is 1) you have a broken application that cannot take advantage of multiple systems and so you need to build it a big old E10k pile of iron to run on (don't go Linux in that case) or b) because your management won't accept that Intel-based systems are "enterprise ready" (in which case they won't accept Linux for roughly the same non-reason).
If you're looking to build out high-capacity infrastructure you go with a large farm of identical 2u Linux or BSD systems (e.g. VA/Linux' FullOn 2xxx series which I have had nothing but good luck with) and you parcel them out as needed. For storage, you probably go with Network Appliance who have the baddest network-attached storage in the business (you'll never go back after taking your first snapshot... 200G of disk... backup time?... 5 seconds).
So, for the home enthusiast Sparc loses. For the the business user this is not even a question, as your constraints are likely pretty simple.
So RH didn't sell that many copies. Oh, waaaah! As if it costs them anything. (If they've got the machines to do the builds now, then they'll have the machines to do the builds tomorrow. Builds can be scripted, so time = money is irrelevent.)
This isn't about money, folks. If they can't afford to mass-produce, they can always produce in low-volume batches and charge the extra.
So what -is- it about? Beats me. Looks more like a decision made to stir up controversy than one made on sound business or programming reasons.
It's a small world and it smells funny; I'd buy another if it wasn't for the money; Take back what I paid (SoM)
I've only got a lowly little old Sparc5, but I've tried Linux, OpenBSD, Solaris 2.6 and Solaris 7 on it. Guess which one I kept? Solaris 7. It's the fastest of the bunch. They were all equally stable as far as I could tell. I recently got a copy of Solaris 8 for it, but haven't even bothered to install it.
I would conjecture that it is in the interests of Red Hat to leave Sun with an appropriate niche, if they are to forge a mutually beneficial alliance in the future. This is only conjecture, of course, but business-wise it makes more sense for Red Hat to not press Sun on "sacred ground".
Concerning MILO, were you trying to do it from SRM or from the ARC? (AlphaBIOS on later machines)... It is much much easier to get it to boot from the ARC. I have installed it on a multia, an Alphastation 233, an Alphaserver 300, an Alphaserver 4100, and even got to play with it for about 3 days on a new GS140 (8-way goodness).
I totally agree with you about support for platforms other than Intel. Most people have little to no experience with any other platform from the Intel norm, which makes trying to get help with a problem on an Alpha (or Sparc or MIPS or...) a total pain in the ass. Much of the documentation is outdated on these platforms.
However, the truth of the matter in the case of the Sparc version of RedHat is that I am sure it probably didn't sell well, if at all. Most businesses with Sparcs would probably run Solaris on them. I would guess that most of the Sparc Linux users are individuals who either are using Linux on Sparc workstations at work, or are using it on Sparcs that they own and run at home. The number of people running a Sparc in their house is significantly lower than people running PPC or Alpha, and is definitely miniscule in comparison to the number of x86 users.
RedHat is *not* Linux... true... but many people see RedHat as the representation of Linux, especially the PHBs. When they see something like this happening with RedHat, they assume that it is true for all Linux. That is where this is truly a tragedy for the community. RedHat is the leading distribution (usage-wise, I'm not wanting to start a distro flame war) of Linux. Until that changes, RedHat *is* Linux is the eyes of most of the public (that even know what Linux is).
My 2 x 10^-2
wolf31o2 Developer, Gentoo Linux Games Team
There's no culpability, and above all, no one to sue when it all goes south
Have you read the typical license agreements that most commerical software companies put out? They pretty much say that even if the program erases your hard drive, you're just out of luck. The whole "we need someone to sue" argument really is a load of crap.
We have a lot of Solaris machines where I work. The main UNIX guy gave an old pizza box to another one of the IS team to learn about it. He had a hell of a time getting Solaris installed on it.
So I suggested Linux. The UNIX guy overheard that and said that he would take the box back if the guy wanted to run Linux on it. "If you want a linux box, use a PC, that's what it's for."
With that kind of attitude, it's no big wonder that they're not selling very many.
Oh, you mean the same Linuxcare that was GOING to have an IPO in June, but decided not to because of "market conditions"?
If you would have taken, say 30 seconds or so, the time to check any of the brokerage houses out there, you would have seen that many dozens of privately funded companies chose to delay an IPO.
But it doesn't sound like a support structure I'd want to depend on for mission-critical systems.
What difference does it make if it is a privately owned company or a publicly traded company?
{snip] they expect to do BUSINESS with a BUSINESS.
Since when is a private business not a business?
In the immortal words of Socrates, who said; 'I drank what?'
I gereally buy from a resleer instead of Sun directly nad get a better price? Isn't that illegal in 6 states?
"Ancillary does not mean you get to rule the world." --U.S. Circuit Judge Harry Edwards, speaking to the FCC's lawyer
[By the way, the mirrors are here: http://www.linux-mandrake.com/e n/f tp.php3#usparc]
RedHat drops support for Sparc; millions switch to Debian!
On i386, alpha, arm, m68k, PowerPC, sparc, and even Sparc64, Debian is still the breakfast of champions. Once you get past dselect, that is. But they're working on it! That's what apt is for.
Right. Red Hat is simply trying to fill their wide market, they're not trying to fill every niche. There's also no Red Hat support for non-x86 handheld platforms, but that's not stopping Linux from being there.
This happens all the time. With RedHat 6.2 the alpha version was released later than the x86 version. I think most Linux distro's show the same behaviour: develop & release for x86, then patch up & release the "weird" ports. The x86 version is the one that brings in the $$$. Give them some time...
True, true, but once you have it you can install it on as many platforms as you like. And, in Suns defense, there are something like 10-15 CD's in the package - I have it and it comes with two full booklets of CD's, plus assorted multi-linugal install manuals, etc. Not a bad deal.
Judging the entire Alpha line of boxen by it's first incarnation (the jensen) is not a fair comparision. Jensens are totally different from any other Alpha. It's the equivilent of a 286 in terms of age. There is a great step by step howto on installing RH on a Jensen at alphalinux.org, it was just updated recently for use with aboot-0.7 and all supporting files updated as well. http://www.alphalinux.org/docs/jensen.shtml .
All other Alphas (except maybe ruffian) are a breeze to install linux on and the documentation is plentiful.
Regards,
Peter
--
www.alphalinux.org
www.alphalinux.org
RedHat made the decision because obviously spending more time on development for the other platforms that they support is more profittable. When a company makes a stategic move, it isn't a blow to the community, unless it's like, massive layoffs and closing their doors. RH is doing fine...
Eh...
You, and all the people who are agreeing with you are either
- not in the real world
- independently wealth (see above)
- dumb
I say cheers to RedHat for being bold enough to support Sparc in the first place, and again for being smart enough to drop said support. I hope they are able to improve the core capabilities by removing Sparc from their support, design, and testing.Besides, I'm sure that nine out of ten of you whiners don't even own a Sparc and are just upset in principle, that you're no longer getting something for free.
"Don't trolls get tired?"
For a good corporate-suported Free UNIX on sparc, NetBSD + Wasabi Systems http://www.wasabisystems.com/ would seem to fit the bill. I've always had much better results on non-x86 hardware with NetBSD, too. It Just Works [tm] instead of being a constant uphill battle because most kernel coders, distro maintainers, etc. don't use that hardware.
Both SuSE and Mandrake offer Sparc distros, and both, though especially SuSE, offer excellent support for businesses. Linux isn't about RedHat and a motley crew of hacker distro's. There are several large commerical and professional organisations willing to help.
The hardware producer manages to put out a decent OS of their own (Solaris, heard of it?) so its not as if the hardware was descending into the Windows arena no is it...
MSBPodcast.com The opinions expressed here are my own. If you don't like 'em... Think up your own stuff.
Of course NetBSD has supported Sparc for a verry long time and will continue to.
Maybe RH is trying to work some deal with Intel? Maybe it's because Solaris is mostly free and has full driver support for all hardware?
It may not be a major blow, as more and more companies are using Intel processors in their servers (whether this is a good idea is another matter :-)
Can't blame them, since they're generally cheaper...
So it does not say anything about Linux usage, more about Sparc not being a competitive offer (I think/guess)
It is a pity however, the more platforms supported the better, ofcourse!
Every expression is true, for a given value of 'true'
While those people posting with the "but there are other distros..." have a point, they're missing the primary concern. Redhat, actually provides support, which seems to be a sticking point with big business. They want someone to yell at when everything goes to shit....Redhat is that someone.
PRAY FOR MOJO
you have Solaris 8, its FREE up to 8 CPU's, and for SPARC is the best choice.
;-)
Yeah, if you have a Sun4M model CPU or better. Most people out there running Linux and SPARC are doing so to get productivity out of older SPARC hardware that Solaris(tm) does not support. Sun has always had a very nasty habit of dropping old hardware compatibility with new versions of their OS.
Of course is a pity RedHat no longer support SPARC, but like someone already said, you have many other distros with SPARC support, and in my opinion better than RedHat.
Both true and false depending on which OS distro you side with. For my money NetBSD seems to run much faster than Linux on older SPARC hardware and Debian has broken pieces but still holds together despite all.
Linux in general is probably more "app" compatible than most BSD based 'nix. (Caveat: That last statement may not be true now, I haven't looked recently - no flamebait meant.)
Regardless Solaris is not the choice for older SPARC hardware - unless you never want to be up to date...
The Tick - "Spoon!"
NEO - "There is no spoon."
"Bah!" - Dogbert
The issue is far more complicated than what either of you make it out to be.
These benchmarks: What exactly are they measuring? No information is given about the hardware other than it is a 50 Mhz system.
My suspicion is it's a Sparc 10 with an SM50 processor card?
If so, the benchmarks aren't measuring 64-bit optimizations that exist in either Solaris or Linux. Perhaps on the UltraSparc 10 this other person has, Solaris does run better.
The benchmarks do clearly show that disk access is faster under Solaris than Linux. This is a very important aspect of total system performance.
What about dual-processor performance? It's relatively easy to add an extra SM50 processor to that Sparcstation 10. How does Linux fare compared to Solaris in that capacity?
I also read the comments on why linux is faster than Solaris. I wouldn't be too sure of that either, over the years the Linux kernel has become more and more complex and has adapted a lot of the same overhead that exists in commercial OS to solve the same problems.
Over time, who is to say that they won't look the same just Linux taking the long road to get to the same point?
The article's subject should be "Red Hat Linux 7.0 possibly won't be released for sparc".
The sparc machines are still part of our build system, and we won't just drop it off.
If we ever decide to discontinue sparc support commercially, the sparc port will be turned into a community effort.
This message is provided under the terms outlined at http://www.bero.org/terms.html
People should keep in mind that RedHat 6.0 didn't support Sparc or Alpha (I don't remember which) either. They didn't have the time to get everything cleaned up and ready for the .0 release. But Sparc/Alpha support reappeared in 6.1, after all the other changes had time to settle.
Red Hat is one of the best companies I've ever seen in terms of not taking advantage of anyone. They really and truly do have their customers' best interests at heart, and I think for that reason alone they will ultimately be wildly successful.
The conspiracy theories are just flat unlikely. It's a HELL of a lot more probable that they stopped supporting the platform because there were about 12 downloads of Sparc Redhat, and those were all the mirror sites.
It's not a popular platform for Linux. Aren't they being smart by re-allocationg their Sparc guys to some other project that will benefit more people?
Apparently no one wants support for the SPARC version. So why would any company want to provide support for a product that has no market?
RedHat decided to do this so they wouldn't have to throw money away, they do have a commitment to their shareholders now to make money. You know, if you see this as an un-claimed market, you could start your own company and support a SPARC version of Linux.
But, alas, it's a bad choice to pursue linux on SPARCs.
I'm guessing the major demand for a SPARC version of linux comes from people who have aquired old sun boxes and just want to fool around. These people don't purchase support.
But, keep in mind Linux != RedHat. RedHat may be considered the premier distro for companies who are looking for support from the vendor, but Debian, SuSE, et. al. are in many folks minds (and for many purposes) as good if not better than RH.
So, this is not a blow to Linux. If anything it should allow RH to put more effort into their x86 distro.
"but I'm sure there are a lot of people who appreciate having a distro they can stick on some old SparcStation."
Which is precisely the reason they're dropping it. Of the small fraction of Sparc machines running Linux instead of Solaris or the ancient SunOS, I'd hazard a guess that most of those are old recycled hardware.
The open-source aspect of Linux is certainly nice given Sun's tendency to leave certain Solaris bugs unfixed year after year. But few people are going to buy new Sun equipment and reload it with an OS that makes mediocre use of the latest hardware.
No, I suspect Sparc Linux is what gets loaded on machines too old and underpowered to run the recent versions of Solaris well. Linux is a good way to get a modern, up-to-date *nix running on all those old 125 MHz Turbosparcs.
And if you're still updating 125 MHz Turbosparcs, you're probably not spending money on packaged OS media and support contracts.
For another thing, why run Linux on new Sun hardware? You're paying a premium for hardware that doesn't run optimally under Linux. Linux can't do SMP as well as Solaris, and you're giving up things like hot-swappable processor support and so forth. Next, you're using this fancy hardware to run what? Apache, MySQL and BIND? It would be nice to be able to run big, scary software like Oracle, DB2, Domino and WebSphere on that fast hardware, but none of those commercial apps are available for Sparc Linux.
Phooey. If you run Sparc Linux, you're unlikely to have been paying for support anyway. Why should RedHat pay people to build and tune Sparc versions of their distro? You're free to grab the RH7 sources and compile them yourself, or put together a team to do so, or switch to something like Debian, Mandrake or SuSE.
For an example of a low-demand OS's demise really disrupting things, consider OS/2's death, officially taking place in the summer of 2001. It may not be in wide use on desktops or even on network servers, but it's got a large market share in the voice-mail systems industry.
Do you ever notice how everyone thinks their own needs are the most important?
Redhat is obviously not making any money from their sparc version, and is making the smart decision to drop it. If people were buying it, they would certainly keep producing it, it's a business decision, and a good one.
My question to the person submitting this article is, did you buy the sparc version, or did you download it?
________
Consider this: With the exception of those of us with old Sun IPX boxes lying around (*grin*), Sparc customers tend to be people who need "big iron." I.E. 8, 16, or more CPUs and oodles of memory.
:>)
Solaris is well tuned to this sort of hardware environment, and the needs that produce demand for that kind of hardware. Linux still has some deficiencies in this area. Consider Linux's 1-1 thread/process model. Linus considers this to be a Good Thing(tm) and just wants to reduce the cost of a context switch as much as possible. Cetainly admirable but Solaris's 1-m thread/process model tends to be the Right Thing(tm) for this type of environment. Our MySQL server running on a Solaris E4500 is serving 4,500 active connections at any given moment during peak usage hours. It would probably need more but that's what we have our MAX_CONNECTIONS set to right now. We're considering upping that as high as 6,000 in the not too distant future.
Between the relatively high cost of a thread on Linux (in terms of spawning and context switching primarily -- I'm less concerned about the 16KB or so of overhead per thread since that's actually quite reasonable) and the remaining coarse-grained kernel locks, plus the somewhat flawed I/O in 2.2 (ladder-I/O starvation anyone?) Linux isn't quite ready to handle 8+ CPUs in a high demand environment just yet.
But even when Linux *does* get to that point (and it *is* getting there -- the kernel team is doing an incredible job with 2.4, and with companies like IBM helping out at the high-end they certainly have a lot of information at their disposal...) you must consider that usually Sparc hardware comes as a bundle with the software and support. This provides a strong disincentive to use Linux -- you've already paid for a generally excellent OS and top-notch support for it.
This tends to relegate Linux/Sparc to older castaway Sparc hardware which has the distinction of not being cost effective to keep up anymore. Performance vs. space and the cost of maintaining one more box (read: TCO) tend to make it cheaper to replace two of those old Sparc machines with 1 brand-new Intel machine.
Personally, I'm more interested to see Linux on IBM's NUMA-Q based machines with 64 processors... IBM is investing a HUGE amount of energy into making Linux *the* OS for such machines. *drool* (Yes, I'm aware that Linus thinks NUMA is fundamentally the Wrong Thing(tm)... It's still VERY fast though...
-JF
MrJoy.com -- Because coding is FUN!
"Not to mention that they have the support in place to handle what large companies demand."
And ONLY what large companies demand. Of those large companies using Linux, how many are using it on Sparc? Few to none, is my guess.
Furthermore, it's easy to see why. Linux on Intel is easy to install and find support on the Internet for. But I've done a few Linux on Alpha installs, and let me tell you that once you leave the warmth and light of the x86 world you are on your own. This isn't to say that AlphaLinux is no good--far from it. I like the hardware and the software. But the support is hard to come buy--if the mailing list doesn't respond or doesn't know you are pretty much toast. Sparc Linux, being even more rare is probably 10 times worse.
I'm sure to be marked as "flamebait" unless I include some examples, so here we are:
Installing on a Jensen.
Using MILO (I've read and re-read the howto and damned if I can figure it out)
The many many (many) patches and updates you need to install after getting a distro (say, RedHat) installed (system clock date to 2020, net-tools, etc).
Again, I'm not saying the above problems make Alpha Linux bad--I'm saying that the poor documentation of the above problems makes Alpha Linux scary.
--
Non-meta-modded "Overrated" mods are killing Slashdot
(Hey Ryan! Here's your proof!)
I suppose this isn't the time to mention my own preference for Debian, which has supported sparc for a while and will continue to do so...
If you are modding me down because you disagree with me, use the "Flamebait" category, not the "Troll" one.
I run a few machines with sparclinux, and have been involved with it for quite a few years (as a lurker). Its painfully obvious that SUN don't provide the same levels of support for their hardware that you can find from any other vendor (eg. Digital{now Compaq} providing Linus with an alpha, coding efforts within IBM and SGI...)
Yet, SUN are happy to gain news inches on the back of linux.
Looks like I'll have to move to debian or suse though...
and of course, if the sparc based machines have to move then my intel ones will have to follow.
-- Don't believe everything you read, hear or think
Wonder What IBM Had to Do With That?.....Seeing as how *they* compete with Sun alot more than Redhat does, and IBM is a major partner with Redhat.
I'm still working on a clever footer.
That's really unfortunate, because my primary computer at work is a Sun Ultra 1 running 6.2 and I absolutely love it. Once you take the hour or so to tie it down (change out wu-ftpd, close up inetd, install ssh, etc. etc.) it's pretty damn cool. Even X is pretty speedy on it. It's by far one of the best "work computers" I've ever used.
Oh, by the way, I'm also a Windows NT administrator.
Too bad Red Hat dropped the ball.
I've been using Sparc Linux for about 5 years now, and Red Hat bringing out a Sparc distribution was one of the things that legitimised the platform in the early days. Now, of course, there's Debian, SuSE and probably others, but I like Red Hat. I'll be sticking with RH on my Sparc for now, with kernel upgrades as appropriate. As major upgrades become necessary, I'll probably recompile stuff myself from now on. It's not really surprising. Everyone that has a Sparc has a lifetime license for Solaris, so the incentive to switch to Linux is much less than for other platforms, and there really isn't a good business case for continuing to support Sparc. I wonder how that'll affect RH emplyees like David Miller and Jakub Jelinek, who still use Sparc as their primary development platform, AFAIK. Obviously, the Linux/Sparc effort isn't going to die, and Sparc support in the kernel will continue indefinitely.
"The invisible and the non-existent look very much alike." -- Delos B. McKown
(1) RedHat isnt the only Linux based OS on the market that has/does support Sparc.
(2) Linux isnt the only kernel on the market that supports Sparc.
What about Debian? What about all the other Linux distros that run on sparc?
Both the NetBSD and OpenBSD ports to the Sparc architecture are quite good.
If you have a sparc, I would think you would be running something other than RedHat anyway.
Just my 2 cents.