A new job opening posted on Sun Microsystems’ Web site confirms the company’s growing aspirations to roll out servers based on AMD’s Opteron processor. “Solaris is the world’s premier 64-bit operating system,” the post on Sun’s job board reads. “Now Sun is taking the next step: to make Solaris the operating system of choice for the AMD64 architecture.”
Sun to Port Solaris 64bit to AMD Opteron?
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
but what if, and I suspect it could happen, a Solaris Opteron solution is tremendously faster then a Solaris Sparc solution? We shall see….I wonder how long the porting process would take, I wouldn’t think too long…
This is great news! The Solaris for AMD64 will waste the SPARC version. Don’t get me wrong, SPARC is an open architecture but development has really languished. I hope they get the beta test out soon so that users can start testing it heavily. I have a feeling that Solaris on AMD64 is going to be a very good performer.
I rather like the Sparc processors. I’m not certain whether an Opteron — at the same clock speed — would be faster.
Then again .. Sun really does need to consolidate sometimes. By far one of my favourite companies, however it’s future isn’t looking very nice
.
This is great news! The Solaris for AMD64 will waste the SPARC version. Don’t get me wrong, SPARC is an open architecture but development has really languished. I hope they get the beta test out soon so that users can start testing it heavily. I have a feeling that Solaris on AMD64 is going to be a very good performer.
SPARC is being developed, the multicore is in development but unfortunately by the time it does appear IBM would have already had its Power4+ multicore out for almost 18months to 2 years.
As for killing SPARC, depends. Most of the server volume these days have been on the 2-4-8way front. Anything about 8 way is still considered a niche and is dominated either by IBM or SUN, depending on which way the wind blows.
What I would love to see is SUN get rid of their Xeon offeres and start selling Opteron equiped servers with Solaris 32bit loaded and offer a free upgrade to the 64bit once it becomes available. IMHO, it would really push their server sales up, and provide all the perks of 64bit computing at a lower cost.
Also, with Opteron running a well respected enterprise class UNIX such as Solaris, it would give a huge boost to AMD in their push into the server market.
I couldn’t agree more. Dumping their Intel dependency can ONLY do good things for SUN and AMD. This needs to happen now, the longer they wait the less effective it will be!!
This could be a win-win for Sun and AMD. Though it’s drastic, I think Sun should stop developing SPARC and run with Opteron – a high-performing, 64-bit platform with hopefully a bright future. It seems that SPARC doesn’t cut it any more, and with Sun being in the situation they are, they could cut a lot of cost out by killing it.
I would certainly be interested in such machines, if they ever come into existence. Our current big Sun machines seem incredibly expensive and not powerful enough; at least not in the pure CPU department.
This could be a win-win for Sun and AMD. Though it’s drastic, I think Sun should stop developing SPARC and run with Opteron – a high-performing, 64-bit platform with hopefully a bright future. It seems that SPARC doesn’t cut it any more, and with Sun being in the situation they are, they could cut a lot of cost out by killing it.
I wouldn’t call it drastic, I would just call it common sense 😉
SUN only has three options:
1) Drastically cut their prices for servers to the point that performance vs. price they’re equal to x86
2) Maintain the price but pump up the number of CPUs in a box
3) Drop the SPARC and replace it with Opteron.
SUN has the engineering capacity to bring many of the highend features over to AMD and with SUN’s cash reserves, they (SUN) are in a good position to buy a large stake in AMD.
Yayy…another one migrating to AMD64!
This is where AMIGA OS needs to be. I’ve said it and said it…
Looks like the 64-Bit OS exodus & subsequent wars are beginning.
–EyeAm
I’d prefer Sun also holding on to their Sparc department. I have way more trust in Sun hardware than any x86 equipment. However I think it’s nice to see they keep offers coming for AMD and Intel CPU’s too.
Sun’s future looks indeed very promising no matter what many holocaust people are saying. This is another piece falling into place in their competitive offer. Good job Sun, it’s comfortable being your client =)
It is time they realign their oss startegy. If only i could get my hands on the source these new hires will be working on ! No one can afford a sparc but everyone is going to get an opteron machine for his home use very soon. They should at least get it to the xen team ( http://www.cl.cam.ac.uk/netos/xen ) so we can get solaris and linux running side by side.
So now you can choose between Java Desktop System (MadHater) of Solaris for your brand new AMD64 box. Great!
JDS and Solaris both use GNOME. Great!
But will Java Enterprise System (project Orion) run on MadHatter? They told it’ll run on Linux and Solaris.
There’re so many talking about Sun having no future…I think there is a bright future for Sun… (JDS, JES, AMD64, …)
offtopic: have a look at the source of the hme linux drivers in the 2.4 kernel!!!
Who will dominate computing in the forseeable future? Intel and Windows and some Linux in the backroom. Intel currently enjoys 50% profit margins. AMD is small, losing money and has not yet been able to balance the field (because of FUD?) Sparc will dwindle. Why would anyone buy it? It has no religious following as Apple/PPC. What should SUN do with its microprocessor fabrication capacity? It could start another line of x86. Or join AMD (and IBM). Solaris already runs on x86. And port Solaris layered products to also run on Linux as purchaseable add-ons.
many sugest that sun should stop with sparc and switch to opteron. i just have one question can juh scale the opteron
in the same way up to 106cpus like in the fire 15k.
ofcourse for a 2-8 way machine it maby would make sense
but what is your opinion would the opteron be a good cpu in bigger machines
What should SUN do with its microprocessor fabrication capacity? It could start another line of x86. Or join AMD (and IBM).
Umm, Sun doesn’t fab their own processors, they outsource fabrication to Texas Instruments.
A fabrication facility is a multibillion dollar investment, and one Sun has not seen as beneficial enough to outweigh the initial cost.
Also keep in mind that IBM and Sun are competing in the realm of server hardware…
I though that Sun would use Linux with the Sun Desktop on a possible AMD64 desktop. It could run many Linux desktop applications and Linux could get some kernel development help from Sun …
Seems like AMD64 will be only for servers and Sun’s server software ?
When AMD64 get full 64 burts (2004?) it maight be as good as the Sparc CPU, if not better – 90 nanometer and lower chip fabrication will be key to more CPU efficiency.
Can the Sparc fabrication match this ?
At some point, benchmarks like these are going to hit the mainstream media and cause quite a stir.
Opteron 1.6Ghz, 64-bit Build
verify/s
rsa 512 bits 39091.0
rsa 1024 bits 16773.5
rsa 2048 bits 6073.6
rsa 4096 bits 1901.3
verify/s
dsa 512 bits 4304.4
dsa 1024 bits 1770.8
dsa 2048 bits 593.9
SunFire 4810 900MHz Ultra3
verify/s
rsa 512 bits 6570.2
rsa 1024 bits 2116.5
rsa 2048 bits 588.8
rsa 4096 bits 164.0
verify/s
dsa 512 bits 535.0
dsa 1024 bits 172.5
dsa 2048 bits 49.3
Sources:
I performed the 64-bit opteron openssl speed test after building openssl for x86-64.
I got the Sparc3 score here http://www.macosxhints.com/article.php?story=20020113045343563
How does your box compare? run ‘openssl speed’. You’ll see the above results in the last column of the last batch of output.
kevin
PS – (just image what the FX 2.2Ghz could do
What should SUN do with its microprocessor fabrication capacity?
I picked this, among your other bunch of nonsense. Just for your information, Sun is fabless. TI is fabbing their CPUs.
You can buy SPARCs that are over 1 GHz in speed, not to mention that with RISC architectures, speed vs. GHz is better than with CISC (x86). There are Ultrasparcs that do 1.2 MHz and faster, and there are the SPARCs that Fujitsu makes, these are 1.4 GHz and more. How come Fujitsu makes SPARCs, you ask? Because it’s an open architecture.
sun had spend lots of money optimizing solaris for x86 over the past decade it could be wonderful operating system, better than linux!
Sorry for breaking it up to you like this, but Solaris IS already better than Linux, and the margin is actually increasing.
P4 2.4Ghz (800Mhz FSB), Debian 0.9.7c binary
verify/s
rsa 512 bits 11560.2
rsa 1024 bits 3969.7
rsa 2048 bits 1179.7
rsa 4096 bits 326.6
verify/s
dsa 512 bits 1059.5
dsa 1024 bits 347.4
dsa 2048 bits 101.3
Have you seen the FX-51’s SPECmarks?
1282 int/1371 fp w/ gcc on Linux
1376 int/1329 fp w/ Intel 7.0 on XP
This bodes well for faster releases of the Opteron… the only cpu that can best these numbers from a quick skim of the results is the Itanium 2, especially in its monster 6MB config w/ Intel compilers
I’m waiting to see somebody benchmark the Opteron in 4/8 way configs against its competitors.
So my 1.7 GHz P-IV outperforms a 900MHz Ultra3 and is blown away by an Opteron 1.6Ghz?
sign verify sign/s verify/s
rsa 512 bits 0.0016s 0.0001s 623.5 8002.6
rsa 1024 bits 0.0078s 0.0004s 128.9 2600.1
rsa 2048 bits 0.0453s 0.0013s 22.1 766.2
rsa 4096 bits 0.3100s 0.0049s 3.2 203.0
sign verify sign/s verify/s
dsa 512 bits 0.0013s 0.0016s 773.7 639.7
dsa 1024 bits 0.0039s 0.0047s 255.6 211.7
dsa 2048 bits 0.0131s 0.0161s 76.5 62.0
Sun makes great harware (SPARC boxes), great software (Solaris), and offers great service (the place I used to work at had a Sun service contract and were happy with it).
Why are they even bothering with other arch’s? Solaris has been tweaked and re-tweaked for SPARC. It’s a stable working solution that many companies are happy with.
Sounds to me like some bean counter thinks that the quarterlies need to look “even better”(tm).
sign verify sign/s verify/s
rsa 512 bits 0.0014s 0.0001s 724.9 7280.4
rsa 1024 bits 0.0085s 0.0005s 118.1 2107.7
rsa 2048 bits 0.0576s 0.0017s 17.4 584.3
rsa 4096 bits 0.4079s 0.0065s 2.5 153.5
sign verify sign/s verify/s
dsa 512 bits 0.0014s 0.0018s 710.9 567.6
dsa 1024 bits 0.0046s 0.0055s 215.6 181.4
dsa 2048 bits 0.0166s 0.0207s 60.3 48.3
Out of curiosity I got this on my Athlon XP 2.4 + (cygwin = no reboot)
Does it almost match the sparc 900 or what ?
🙂
Dear Genius,
As I was reading your comment no room for cpu architectures besides x86 I at least took it to be a pessimistic but possible view of the future of standardization much like TCP/IP is not the best protocol but it’s too late to change it. Despite what appeared to be a pessimistic outlook I still respected the comment. Altough I would argue that it wasn’t true since x86 is an UGLY architecture (I’ve programmed both x86 and SPARC assembly) and it’s ineffecient (requires CISC->(internal)RISC translation) and the fact that technology always advances. It is a skill to hoist an engineer on his own pitard. You did it yourself with this stroke of genius:
they should go with intel itanium instead, because it is full 64-bit processor instead of pseudo 64-bit like optoron
It all makes sense except for that an Itanium is *NOT* an x86. It is an EPIC processor. So not only is it not the same architecture as an x86 (CISC (w/ RISC core in the modern implementations)) it is a COMPLETELY DIFFERENT DESIGN PHILOSOPHY. The merit of the Itanic is questionable but the validity of your post is not.
And on a final note the SPARC is the best at what it does: scale. Which is *EXACTLY* what you expect a Scalable Processor ARCitecutre to do. Comparing SPARC to x86 is apple to grapes.
An Alpha 21264, 800 mhz, if people are curious about having another architecture to look at:
sign verify sign/s verify/s
rsa 512 bits 0.0009s 0.0001s 1123.9 13614.1
rsa 1024 bits 0.0027s 0.0002s 368.7 6188.7
rsa 2048 bits 0.0141s 0.0004s 71.2 2266.9
rsa 4096 bits 0.0856s 0.0014s 11.7 712.6
sign verify sign/s verify/s
dsa 512 bits 0.0005s 0.0006s 1849.7 1665.6
dsa 1024 bits 0.0013s 0.0015s 779.9 652.7
dsa 2048 bits 0.0038s 0.0045s 265.8 223.2
OS is FreeBSD 5.1
Looking at their job post page they have offices in countries all over the world. They are developing Solaris for x86, x86-64bit and SPARC CPU now, it doesn’t look tobe dying to me.
Looking at their job post page they have offices in countries all over the world. They are developing Solaris for x86, x86-64bit and SPARC CPU now, it doesn’t look tobe dying to me.
Well, there is the issue of the recent layoffs and dips in their stock price due to their recent earnings report…
However all of this must be taken with a grain of salt… Sun has been retooling all of their product lines across the board with the Mad Hatter and Orion projects (which now form the Java Enterprise System)
Sun should be judged on the success or failure of the Java Enterprise System, and also on their traditional sources of revenue (hardware/software sales, service contracts), not on their recent development losses before launching a major new product/service.
This isn’t about killing off SPARC.
This is about killing off Itanium in the low-end of the server market.
The single-cpu and 2-4 way server.
Sun sees the Itanium as a future threat to their “big iron” systems – the 32way-plus monsters.
But right now, Itanium is trying to get its foot in the door at the low end, so it can get the software ported, etc.
If Sun can block this by offering AMD64-based systems at a reasonable price – and remember, AMD64 has top-class x86 performance as well as the ability to run proper 64bit software – then they can try and stop Intel getting their foot in the door, so to speak.
IBM may adopt the AMD64 for similar reasons, or they may push low-end PPC970-based machines. Or even both. IBM is huge and has many divisions, after all.
Either way, it’s in both Sun and IBM’s best interests to block Itanium – and thus its supporters, HP and SGI.
HP’s strategy seems uncertain right now – tho’ the PA-RISC -> Itanium wasn’t exactly unexpected, despite a previous poster. In fact, Itanium can be considered the natural successor to the PA-RISC. But HP seems unsure what OS(es) they should push on such a system, and seem bizarrely enamoured with Windows Server. SGI are all but dead at this point. They’ve not stopped twitching yet, but it’s a matter of time. A shame, and they will be mourned, but that’s how it goes.
Of course, the big winner here is likely to be AMD, since they’re gaining adoptions for their new server by industry stalwarts. The big loser, of course, would be Intel, given how much they’ve put into Itanium. And how much they’re still putting into it.
Yes, we live in interesting times…
many sugest that sun should stop with sparc and switch to opteron. i just have one question can juh scale the opteron
in the same way up to 106cpus like in the fire 15k.
ofcourse for a 2-8 way machine it maby would make sense
but what is your opinion would the opteron be a good cpu in bigger machines
As much as some people would like it, SUN’s recover will not be on the back of 106 way SPARC machines. Their recovery will be based on an increase in software and x86 hardware sales.
Look at Itanium for example, in the couple of courters that AMD Opteron has been on sale, AMD has beaten Intel in terms of the over all number of Itanium CPU’s that have shipped.
As I said previously, 2-4 and to a certain extent 8 way machines are the market sweet sport, only a very, very small amount goes to configuratios above 8 way.
As for Opteron scaling, there is nothing stopping it from scaling to great hights, Cray for example created a super computer based on the Opteron.
The thing is, AMD is focused on volume and marketshare. It is a better investment for them to concertrate on what the vast majority, 80%, want rather than serving the ultra niche market who want 100s of CPUs in a box.
The fact is, SPARC isn’t dead but developing too slow, had their multi-core CPU came out at the end of this year, beginning of next then the possibility of a good recovery would be possible, however, when you’re pushing out under performing CPU’s, it is pretty logical to see what the outcome will be.
Even if Opteron did perform like that, one could easily purchase a Sun Crypto Accelerator Board from SUN which off loads the type of thing on an expansion card:
http://www.sun.com/products/networking/crypto.html
SPARC should still obviously work past 8 way systems, but from 1 CPU – 8 CPU systems they should exclusively use the Opteron for all Workstations and Servers. By using larger numbers of CPUs they can lower the costs and increase yield efficiency. They would also be able to stay competitive in the x86 market and not hand the money directly to Intel. Further they can hit fast and possibly beat Windows AMD64 before it gets too much traction. They may also want to consider PowerPC for 16 way and beyond if their own SPARCs can’t handle it within the next year, it shouldn’t be that massive of a change, they are both RISC chips and have much more in common with each other than x86 or EPIC. It would certainly be nice to steal away some of those AIX and maybe even Z/OS customers.
Is the Itanic quite the quagmire for Intel that the industry press has portrayed it as? Any one have any experience with it? From what I understand they off-loaded all the complexity on to the software… and the software can’t exactly manage. I assume the issue is compiler performance?
dual Xeon2.4 w. RH7.3 (xSeries 235)
sign verify sign/s verify/s
rsa 512 bits 0.0012s 0.0001s 829.8 12500.3
rsa 1024 bits 0.0050s 0.0002s 201.0 4035.6
rsa 2048 bits 0.0304s 0.0009s 32.9 1099.0
rsa 4096 bits 0.2147s 0.0035s 4.7 285.8
sign verify sign/s verify/s
dsa 512 bits 0.0008s 0.0010s 1189.1 967.1
dsa 1024 bits 0.0026s 0.0031s 380.7 325.9
dsa 2048 bits 0.0091s 0.0115s 110.2 86.6
Is the Itanic quite the quagmire for Intel that the industry press has portrayed it as? Any one have any experience with it? From what I understand they off-loaded all the complexity on to the software… and the software can’t exactly manage. I assume the issue is compiler performance?
In theory, Itanium is a great processor. A clean ISA, built from the ground up with the latest and greatest concepts taken into account. In practice, however, people see it as nothing more than an attempt by Intel to monopolise the market by having a processor which is produced by one vendor.
I have used an Itanium II equiped workstation and yes, they are VERY nice, however, people aren’t going to adopt it when they charge the price they do. Worse still, Intel just doesn’t get it. Using New Zealand as a generalised benchmark, almost 1/2 the computers (desktop and server) sold in New Zealand are from non-multination companies. They are locally owned companies.
If Intel want to push their Itanium, they need to open the Itanium market, and get third party motherboards on board AND start distributing their CPU’s through distributors such as Techdata Pacific, VST, Melco etc etc.
When I had a reselling contract with Melco, I could source almost any processor I wanted except the Itanium. Itanium will not take off until, as I stressed, the price is reduced and there are third part motherboard producers get into the market.
The second part of the “prong” is software. Intel has the cash. They need to push the workstation and desktop so that the price can be reduced due to economies of scale. How do you push the workstation and desktop? get the software producers on board. Approach Microsoft, Macromedia, SGI, Adobe, Quark, Corel, Borland and various game producers and ask, “How much will it cost to port to IA64”, find out the price, cut a cheque.
Once the software base is there and the hardware is priced nicely, then you will see people jump on board. Until such time, people will continue to call it, “Itanic”.
Just to add to the previous post, a point I think is curcial: there’s no software for the Itanium, and while it’s able to run legacy x86 code, it runs so pathetically slowly (PII-233-300 MHz) that nobody will want to shell a couple grands for such a CPU to run their current software. So, with Itanium, you MUST port your software. And while the Itanium’s ISA is free of the x86 clutter, it brings a LOT of it’s own freaknesses, and is regarded as a very difficult CPU to program for.
If SUN enters some kind of partnership with AMD it would do both companies good. Sparc is a decent architecture but SUN isn’t keeping up with the rest of the processor world in this area. I am not a huge fan of x86 solaris especially when you have Sparc Solaris however x86 64-bit AMD Optimized Solaris is a differnet animal. Hopefully SUN can capitalize on a good idea and not have Mcneally Screw it up for them.
re CooCooCaChoo
You are so right! It would be nice to be able to order an Itanium Motherboard/Processor for a white box build. The same goes for the IBM processors and Transmeta CPU’s. Building back end server systems doesn’t require a certain Processor type unless you have a homogenous Desktop/Server environment. Wouldn’t be nice to be able to assemble a scratch build itanium server clone to run BSD or linux? rather then having to buy from HP or whoever.
Sorry guys, but wasn’t Sun doomed, dead, and gone forever just a few days ago? Why are you all suddenly writing about the company as if it was alive? Things sure change fast here on OSNews. =)
Don’t think just because a technology is good it will be prosperous. Look at Tru64, hands down the best technical Unix (if you don’t agree you’ve never really used it). Look at AlphaServer, nothing crunched close since it was released in the mid-early 90’s. Way ahead of what Sun was even dreaming at the time. Look at VMS, probably the best OS ever written and still produced, yet HP is letting it drift away. It’s about money and with Sun’s stock looking the way it is, the future doesn’t look good.
Dear Genius CaptainPinko,
I have to agree that the original poster was a moron. But may I also point out that there is a small mistake in YOUR post:
The Scaleable Processor ARChitecture (SPARC) is named after the (theoretical) ability to expand the register stack (up to 512 registers/ 32 windows).
PS: and by the way, it’s called the “Itanium”, not the “Itanic”.
These grand prognostications from the infinite wisdom of Wall Street tend to become self-fulfilling prophecies. It used to be OK for a corporation to lose money to get some back later, in other words eating up short term profits for long term viability. Not anymore, if your PROFITS are down by a penny they drop your stock like it was junk. The market is truly sad, everyone has a day trader mentality, it is all about quick money and not about investing in a company you believe in. Just another game created to make money, like gambling, but with other peoples lives.
I am sure once everyone thinks that SUN is dead, of course no one will purchase anything from them because “hey, I heard they are goin’ under”. Once customers stop purchasing because of these grand mystic forecasts from the great beyond, SUN will be in serious trouble. Appearance is always more important then reality. SUN really should spend a billion of there reserves on marketing, they really need to let everyone know why they are around and what they can do for business.
If everyone hears negative comments about SUN and hears nothing positive in return, people’s perception of the company is going to make a turn for the worst (and it already has). Look how badly Novell was slammed because they were unable to change the perception that their software was “archaic”, even though they were way ahead of Microsoft at the time.
Will Sun simply port Solaris to an x86-64 PC platform, or will they do it “properly” and create a new platform based around the Opteron processor ?
“just what sun needs … another iron in the fire”, i thought that was a serious comment. i also thought it was enough for people to think about, without a bunch of redundant text. i wonder why that message was deleted.
some might remember that lack of focus was a topic in that analyst’s critique.
Will Sun simply port Solaris to an x86-64 PC platform, or will they do it “properly” and create a new platform based around the Opteron processor ?
What do you mean by “done properly”? Solaris for x86-64 will be a Solaris SPARC 64bit to x86-64 port, meaning one would retain all the 64bit features that make up the SPARC version. I would love them to make their OpenGL available as well 😉
Considering Linux and the *BSD’s all have x86-64 ports for a while now, the Solaris port can’t be too hard.
Solaris was sanitized for x86, SPARC and PPC a while back so the arch dependancys in the code shouldn’t create a large problem; usually the hardest part of a port.
It WOULD be interesting to see if the old PPC code could be tweeked to run on a Mac G5 for giggles in addtion to x86-64.
“PS: and by the way, it’s called the “Itanium”, not the “Itanic”.”
FYI the reason people call it Itanic is because it’s sinking like the Titanic.
I think so.
I would much rather see Linux concentrate on the desktop market and leave at least one proprietary UNIX version alive for servers
Solaris isn’t an old nag OS, it has a fully pre-emptible kernel with 64 bit code in mind (since quite a bit of the kernel was re-written to accomplish this.) The threading model is much more mature than Linux and there are numerous other technical superiorities (how it handles memory and I/O, driver modules, etc.)
Even Linus mentioned that he never intended for Linux to be on the high-end servers, he is still even to this point specifically focusing development on workstation class systems, not servers. Although that is not the case for IBM, they are pretty much dumping their AIX (which is better than Linux currently) and pushing hard to enterprise Linux so they can catch the Linux love. (Wall Street loves Linux, it is going to magically make IT profitable again)
Solaris is currently cheaper for corporations in 2 way configurations…
Description: Solaris 9 Workgroup Server2 RTU License, 2 CPU Maximum, x86 Platform Edition
Part Number: SOLII-090-B999
List Price: $250.00
RedHat and SuSE cost about $750 for a year of maintenance and the software. The server hardware we purchased has support for specific Operating Systems, in specific HP is supporting only Server editions of RedHat and SuSE (UnitedLinux) on future releases. Incidentally, HP’s Proliant line now has Solaris 9 x86 listed under supported operating systems.
Have you ever tried to load RedHat 9 on an HP Proliant DL380? It won’t even boot off of the install CDs! (It can be done, but it certainly isn’t “officially” supported.) If you don’t meet the specifications for the hardware, you don’t have support (or limited support) for the product.
I personally don’t have a problem with loading a personal edition and then just using the versions of all the packages/kernel/libs that I want, but for some reason management wanted the fully tested version backed by the manufacturer for the enterprise. Probably a good thing I guess.
“Incidentally, HP’s Proliant line now has Solaris 9 x86 listed under supported operating systems.”
Thanks, that put a smile on my face. Hopefully what a Sun engineer posted here a while back is true – that the battle has been fought within Sun as to whether it is the SPARC architecture or the Solaris OS that is the true value of the company, and the OS guys finally won. The fact that they now have HP pushing Solaris for them is the kind of proof I have been waiting to see.
What do you mean by “done properly”?
By making a completely new platform using an x86-64 chip, not just evolving the PC line with it’s 20-odd years of accumulated cruft like the ISA bus and the PC BIOS.
In short, build new machines just like their current machines, only using x86-64 CPUs – the way Apple would have done it if they’d jumped ship to x86.
SGI tried to build and sell really expensive PCs while pretending they were something more – and look where that got them.
[i]Solaris for x86-64 will be a Solaris SPARC 64bit to x86-64 port, meaning one would retain all the 64bit features that make up the SPARC version.[i]
Will it ? Or will they just hack the ia32 version to an ia64 version ?
I see the Solaris vs. Linux battle has been raging something fierce lately. So for the Solaris people: Is Solaris on x86 desktop usable by newbies or is it only for the Borg? I don’t care about speed, just user friendliness and a healthy amount of available apps…
Well, the article was initially posted becaues SUN was hiring OS engineers for the AMD64 port from Solaris. So I *hope* so, I pray it isn’t going to be a thunking nightmare like MS Windows is going to be in 64 bit (don’t report abuse, thunking is changing 32 bit data segments to 64 bit or vice versa, this is what was done in Windows 95 for 16 bit applications on 32 bit OS). Even if it wasn’t a “proper” OS implementation, the AMD64 couldn’t perform any worse than the SPARC currently is. That is such a sad statement.
An aside…
I have no idea why SUN would let the SPARC languish the way it has, however it definately has. If Itanium gets any more momentum, than EPIC CPUs will be the server chip of choice. Intel has way too much muscle, there were a couple of comments saying that ONLY Intel is going to produce Itaniums and you are going to be locked in and so no one will choose this platform. Let’s pray that Itanium doesn’t get any more traction.
This type of situation is perfect for Intel, the server market is different than the PC market, they would love to be the sole source of all server hardware. They can push their mobos down everyones throat and we will have no choice once it’s an “industry standard”. Intel still has a major stranglehold on CPUs, I don’t see how this is going to change any time soon unless PowerPC starts beating the Itanium (or a miracle happens and a new SPARC is released manufactured at .9 microns running at 2Ghz.)
Well, I’m installing RedHat Advanced Server on HP Proliant DL-380s and I’m not having any problems. Perhaps you need better admins.
And if you had them perhaps you could forget about those support contracts and the costs associated with them. Linux supports itself when you hire people like me.
Yes, AS does work just fine, funny that…try RedHat 9 (not the advanced server edition
– hence the point of my statement. Only supporting Advanced Server or Enterprise Server – install RedHat9 on your DL380, without first installing 8 and then doing an upgrade, or going through the slowest FTP site on earth for your Linux install source. The RedHat9 CD won’t boot on a DL380 G3, but the AS version sure will…
My point was, HP is not trying to offer support on the server hardware for standard or personal distributions, they instead are targeting server editions to support. Which forces me to use those versions if I want HP support. Of course maybe you being the better UNIX/Linux admin don’t need support from your hardware vendor.
Really? If linux supports itself then we don’t need to
hire admins. If you had any real experience in systems admin,
you would know that not only are IT departments, ENG
departments and even executive staff demanding support
contracts for systems, but other products that we buy
to run on linux many times have a strict support matrix.
So guess what happens when you install Oracle on Redhat9?
It doesn’t work. You get to hack install scripts, and Oracle
will tell you it should work but it’s not supported. That
means no legal support contract. Same goes for tons of other
software.
Ah hiring an admin like you would solve this tiny problem huh?
and support was bad to laughable. I didn’t feel it was good value for money.
My experience with RedHat support – It sucked. They weren’t even sure which modules were compiled into the kernel and which were loaded with the boot image on their pSeries Advanced Server version. I recieved two completely contradictory answers from each technician I spoke with. The real point here is HP will not even attempt to support customers that run non-supported OS versions. So basically I am paying tons of money for a bad support contract, updates and software maintenance, and an OS version supported by the hardware vendor. The point is, that version of Solaris x86 for $250 is starting to look very good to me.
And if you had them perhaps you could forget about those support contracts and the costs associated with them. Linux supports itself when you hire people like me.
So I assume you audit the code, produce patches, recompile, package them and test these packages to ensure that nothing it broken. That is maintenance, not the maintenance that IBM preaches, “bugger off and let us control your computers”.
RE: BFG (IP: —.client.attbi.com) – Posted on 2003-10-07 05:34:22
My experience with RedHat support – It sucked. They weren’t even sure which modules were compiled into the kernel and which were loaded with the boot image on their pSeries Advanced Server version. I recieved two completely contradictory answers from each technician I spoke with. The real point here is HP will not even attempt to support customers that run non-supported OS versions. So basically I am paying tons of money for a bad support contract, updates and software maintenance, and an OS version supported by the hardware vendor. The point is, that version of Solaris x86 for $250 is starting to look very good to me.
From my experience with SUN, you will always receive top quality support. Sure, there is a slight premium but atleast you won’t need to ring up multiple times until you eventually get a support technician who has a clue about the products being sold.
As for Dell, the only thing going for it is their cheap hardware, that is it. Their support is crap and even their front line, sales staff don’t know a word of English. Do you know what the sadest part is? they put on fake accents to appear they’re from the US thus making them impossible to understand.
IBM, I have to agree, they’re exactly what Scott McNealy says. They’ll sell you the most complicated, complex system known to man then ram down their “Global Services” down your throat until you submit and sign a contract.
SUN on the other hand, you buy hardware, they’ll offer a support contract BUT they won’t ram it down your throat. They, unlike IBM, realise that no everyone needs to have the ultimate, uber support, most simply need software support, some need a little technical support via the phone and only a small, microscopic require the “uber, 1hour response time, clear the roads for the SUN rescue team to come through” support contract.
As for HP, what a balls up that operation is. Their PC’s are full of proprietary junk that makes it next to impossible to install any other operating system except Windows, their printers are cheap and the print quality even worse, hence I have moved back to using Epson after having used several HP printers. Their scanners are way over priced for what they can do. The only thing I think that is half decent is their calculator and the remaining battle axes who are maintaining the OpenVMS code even with Queen Microsoft Sycophany, Florina screaming, “windows, windows everywhere and we’ll supply the support!”.
As for Dell, the only thing going for it is their cheap hardware, that is it. Their support is crap and even their front line, sales staff don’t know a word of English. Do you know what the sadest part is? they put on fake accents to appear they’re from the US thus making them impossible to understand.
Can’t agree with that. I used to work for a large Australian Uni and our support from Dell was never less than excellent – easily on par with our Sun support. Of course, we had a fairly large investment in Dell hardware (albeit dwarfed by the one in Sun hardware) so that may have had some bearing.
I have an HP49G and I must admit that I love it very much, it is very dear to me, it is my precious….
Anyway, Compaq servers no longer require the damn Compaq system partition. You can use their SmartStart CD to run the Array Configuration Utility, you can actually enter a real CMOS, or you can even configure the RAID arrays at boot time. The Smart Array controller is one of the best things about HP servers. Their hardware is actually damn fine and I have never had any real problems with it (on the rare occasion something doesn’t work, we pull the redundant part and replace it).
Their support on the other hand, leaves a lot to be desired. I have a large array of disks running on a storageworks SAN, when I called HP about 6 or 7 months ago I was able to quickly reach a very intelligent engineer from their StorageWorks group late in the evening that was able to help me very quickly with an issue that had been plaguing me for hours, he knew his stuff. I was impressed. So far so good.
However, I called again about 2 months ago because I had a simple but important question to ask regarding the same hardware system. After being bounced around on their phone system for about 30 minutes, I reached some guy on the line that didn’t know which way was up. I had to point him to the documentation I found on their website for the hardware because I wanted clarification in regard a maintenance procedure, he said “oh cool, I didn’t know this was here”… I said, great… glad I could help, bye bye now.
I think SUN is one of the few companies that isn’t outsourcing their whole technical support operation over to India. Tech support can be a very valuable thing at 4 am on Saturday. Is their technical support valuable, or has it gone down hill at all?
Hi, may be this is offtopic question, but can someone(s) make clear for me which of these are CISC and which are RISC processors say Pentium/Celeron,Itanium,Opteron,PowerPC (PPC?) and SPARC(UltraSparc)? And what are the advantages having CISC or RISC processor in Desktop systems?
Thanks.
“As for Dell, the only thing going for it is their cheap hardware, that is it. Their support is crap and even their front line, sales staff don’t know a word of English. Do you know what the sadest part is? they put on fake accents to appear they’re from the US thus making them impossible to understand.”
Can’t agree with that. I used to work for a large Australian Uni and our support from Dell was never less than excellent – easily on par with our Sun support. Of course, we had a fairly large investment in Dell hardware (albeit dwarfed by the one in Sun hardware) so that may have had some bearing.
You’re in a large organisation who is going to be a good supply of money. If you were a company who makes purchases infrequently and not always from Dell, don’t expect the same level of service.
Here is a good example, I ring up and ask for technical support, I wanted to know the monitor frequencies so that I can input them for screen settings, I was thrown around 8 different operators, I eventually, after the person stuffed around, “bugger it, I’ll find it myself” and hung up. I then googled the internet for the next 5 minutes and found the information I mean.
If they can’t get the small things right, how on gods green earth are they going to get their enterprise services right. Worse part, they’re now “promoting” their US$99 services deal. Nice, now instead of getting a clueless simpleton over the phone I’ll get some teenage PC-fanboy gracing my presence whilst he shoots himself in the foot when replacing faulty hardware.
As they say, you pay with peanuts and you’ll get monkeys. With SUN, you will always get top quality service, not the aim and miss scenarios that happen when you ring up Dell or HP.
I bought a SUN Blade 100 a year ago, I received MORE support than I expected, I was even given a little help setting up CUPS so that I could use my printer. This person wasn’t even required to go that far. When I received my machine, I was surprised it didn’t come with Solaris media, I then went into Solnet, SUN New Zealand base of operations, and *GIVEN* a free copy of the latest version of Solaris.
*THAT* is what I call customer service. Going the extra mile for the customer, whether they’re a multibillion dollar company or a small business with 5 employees.
I have always found it funny that people here demand cheap but then complain that they receive terrible support. Quality support costs money. When hardware and software reduces in price, something has to give, either the product quality is reduced through a loose Q&A process, less call centre staff to handle support queries, or reduce the amount spent on R&D.
When you purchase an Apple in Australia, you get support and sales staff member from Australia serving you. That quality is paid for via higher prices on their hardware. Dell, on the other hand have cheaper machines, meaning, they need to reduce costs and if that means that they have to put their support staff in the wop-wops where no one speaks English, then that is what they’ll do.
people say sun switches strategies too much. i see this, writing solaris for opteron, as evidence that they are switching from one dual-platform strategy:
linux/x96 and solaris/sparc
to another dual-platform strategy:
linux/desktop and solaris/server
that might give them a clear story they can tell (short term) about why they sell linux for one x86 and solaris for another … but i don’t really see that the strategy can last. if they continue to sell linux/x86 servers, or if they are forced to support linux on x86 servers by competition, it tears their own story down. they end up with as much os confustion as they have now.
in retrospect, they should have kept their *first* dual-platform strategy:
cobalt/linux and sun/solaris.
it still is about branding.
The following are RISC architectures (reduced instruction set computing):
MIPS
PowerPC/Power
SPARC/UltraSPARC
PA-RISC
Alpha
Itanium*
(and more)
*) Well, they say Itanium is a VLIW/EPIC architecture. They are right, of course. Let’s just say that VLIW is in general a heavily modified RISC.
These are CISC (complex instruction set computing):
Pentium/Xeon/Celeron (ix86)
Athlon64/Opteron (ix86-64 or amd64)
Motorola 680×0
VAX
(and much more, usually older CPU architectures)
Actually CISC describes not a specific processor architecture; the acronym CISC was just invented to refer to all non-RISC architectures.
The reason why RISC was invented was to try to increase the speed/performance. After studying the instruction set of CPU s and how they are used in real life (real programmes) they came up with the idea to create a ‘minimalistic’ type of CPU to be able to speed up the programmes.
Why? Well, some few (mostly basic) instructions are actually used most the times, whereas some others were used rarely (usually more complex instructions). This is not really surprising, since these complex instructions are of very specific nature and can’t be used in many places (but are handy when applicable). But just using the simple/basic instructions that are used most often you’ll save transistors and logic AND – that was the important point – you can rev up the frequencies more easily. Being able to increase the frequency of the CPU would give the neccessary speed improvements.
Being RISC usually indicates that a CPU has many registers, few, simple instructions, not many address types and just load&store instructions. All the complex CPU instructions had to be implemented in software (ie. a bit slower), but since they are not used that often – and since the CPU is running at higher frequencies – this shouldn’t matter.
Well, at the beginning the RISC concept worked fine: the RISC architectures, e.g. early Alpha were using frequencies much higher than the old CPU architectures and were way faster.
But at some point in history the old CISC architectures changed many of their inner CPU designs to RISC style to allow them to archive higher frequencies aswell (e.g. the newer CISC CPUs are actually a RISC core but have a translator engine to translate CISC->RISC).
And strangely enough all the RISC architectures implemented many not-really-RISC-like features like out-of-order execution and speculative branch decoupling that improved the speed without having to improve the frequency (ie they were suddendly not focussed on frequency anymore!).
So now all-of-a-sudden the CISC architectures overtook the RISC in terms of frequency (odd enough), and the RISC became better in instruction efficiency than CISC (even odder).
Right now there is no real difference in RISC or CISC design anymore: you can only see a clear difference in the instruction set of the CPUs, but how they work underneith is often quite similar. It is mostly a matter of taste, and I for one prefer a simple and working RISC over a CISC.
Ahh, yes, speaking of taste – I forgot to mention: since RISC architectures are usually newer/younger than CISC, they are usually well designed from the beginning whereas the old CISC have gotten many ‘add-ons’ to their initial instruction set, making CISC architectures usually look more like a patch work than a design.
As for ‘what is better for desktop systems’, there is no real answer. As you see PowerPCs are RISC, but Apple builds quite nice desktop machines with them! RISC are usually used in servers but also in visualization centers and graphic workstations, so they can defintely be used in desktop configurations. And Intel Xeon processors are also used in SMP servers, so it seems that CISC is not only a desktop solution.
Well, RISC seems to have a tradition in the server market and CISC in the desktop, but this is just because of marketing (and pricing), but not because a CPU has to be specifically suitable for desktop use!
(today’s RISC may have a higher I/O thoughput than CISC, thus being able to handle lots of data better. But I’m speaking of LOTS, nothing what you would usually be confronted with by using a desktop system
greetings, Max