Sun Microsystems may be planning to drop its private-label version of Linux for servers, but the company is still on track with a Linux-based desktop system that should hit the market this summer, the company said. However, eWeek says that Sun has no plans to support Linux or Solaris on Intel Corp. Itanium systems, but the company is evaluating AMD’s upcoming Opteron processors, Sun officials said on Monday.
Sun Looking to AMD for 64-Bit while their Linux Desktop is on Track
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
if they’re doing what I think they’re doing… that is, Sun-branded x86 application servers for Sunray environments.
Sun has also developed a Linux version of the Sunray server software. It’s clear they’re aware that the existing Solaris application base is not sufficient for companies to warrant purchasing Sunray networks, so now they’re seeking alternatives.
He doesn’t understand open source and has probably never read the GPL. I’d be surprised if Sun makes a profit anytime in the near future. They have no direction while their over-priced technology slowly falls behind (in terms of both performance and stability).
Hellooooo…. people! Jeeezzzzz….
They have no direction
Sun has a very clear vision: automatic management and distribution of workloads across multiple servers (N1) which provide application services to thin clients (Sunray). The problem is they don’t know how to pitch this idea to their clients.
while their over-priced technology slowly falls behind (in terms of both performance and stability).
I’ll certainly concede the issue of price; Sun lost that battle long ago. However, you certainly must be jesting in regards to stability. Solaris and Sun hardware is some of the most rock solid computer equipment on earth. Sun is still the unbeatable champion of throughput. The I/O performance of Sun servers is astronomical… far more than any x86 solution, and much more than other HPC vendors, including SGI.
Um, SGI, I believe is the leader in I/O performance by a wide margin.
But Sun does have really stable really expensive servers. However their workstation hardware sucks and soon won’t be any better than a 64-bit PC. So much for the most rock solid computer equipment theory. They have that up a long time ago, about the same time they gave up SCSI for their workstations.
IDE IS cheaper. Hehehe.
While true that Sun has some solid IRON, with the release of Hammers not to consumers like us but to the HPC market, I think you can expect to see some serious IO bandwidth.
Other day I was at BioITWorld looking at everything HPC under the sun, funny thing Sun/DELL was no show.
One company Microway was showing a dual Opteron server far beyond anything I could buy. It featured HyperTransport AND Infiniband. An AMD guy was in the booth too.
I would think Sun will do well to use the Hammer.
Um, SGI, I believe is the leader in I/O performance by a wide margin.
Sun has been the throughput leader for quite some time. See http://www.sun.com/smi/Press/sunflash/2003-02/sunflash.20030210.12….
The Sun Fire 15k sports 172.8GB/s aggregate system bandwidth, 115.2GB/s peak system bandwidth, and 43.2GB/s sustained. Its I/O bandwdith is 21.6GB/s sustained.
Perhaps you’re confusing system bandwidth with memory bandwidth, and believe SGI has Sun beat due to NUMA. Unfortunately the system bandwidth of SGI machines is rather pathetic, considering they were designed for CPU bound tasks, not I/O bound tasks.
They have that up a long time ago, about the same time they gave up SCSI for their workstations.
IDE IS cheaper. Hehehe.
Blade 2000 workstations use FC-AL. The Blade 150 is a Sun-branded 3rd party product. It’s obviously not intended to be a high performance workstation. It’s entirely incorrect to say that Sun “gave up SCSI for their workstations” when really they’re just providing a lower priced workstation offering. Personally, I don’t see a market for Blade 150s, except for hobbyists who want a cheap UltraSPARC II systems.
Anyone who is repeatedly performing simple transformations/calculations on extremely large data sets (thereby making it an I/O bound task) will benefit from the I/O performance of a Sun Blade 2000 workstation. We process some fairly large (400GB) atmospheric data/model output, and display it 3D rendered in realtime on Expert 3D framebuffers. The amount of I/O required to process that quantity of model data and in turn feed instructions to a framebuffer is a challenge beyond the I/O capabilities of any PC that’s been purchased here (and believe me, it’s been tried several times)
you are stupid and don’t know how modern pcs work. the only way that could possibly stress a pc is if u were using one without 64bit pci, which would be really stupid. all the things you think are specific to sun computer are NOT.
u can buy a motherboard with 64-bit pci
u can buy a (OMG!) SCSI card that is 64-bit pci
u can buy a (OMG!) SCSI hard drive to, one that is faster than the crappy one in your sparc computer
all of this will be just as good as in the sun, and thanks to the 533mhz fsb of pentium 4 u will actually have more bandwidth than your stinky sparc computer. i think ur problems with pcs are because u buy shitty pc then expect it to run like $100,000 sparc computer. ha ha u really make me laugh. that’s says nothing of the memory. pcs now use ultra high speed ddr333 memory or faster. ur sparc is using crappy anchient sdr memory, and i bet its bus speed is like 66mhz to. pcs always get the latest technology, when older systems like sparc lag far far behind.
the numbers u give are impossible. the fastest fsb is the pentium 4’s and it runs at 533mhz. there is no way to move over 100gbps through a computer. 32bit pci can barely move 1gbps. why do u think internet routers aren’t full computers, they are custom built? because no computer can possibly handle that much traffic, that is why. u are saying u can read from disk faster than the fastest pcs can read from their ram? ha ha that is a major laugh buddy
why ppl are so stupid and shell out thousands of dollars for a slow, expensive sparc computer when a pc can do EXACTLY THE SAME THING for a fraction of the cost is beyond me. people are just dumb and have loyalty to their sinking ships like sparc because they are so incredibly misinformed about pcs.
pcs are the future. sun will be dead in less than 5 years. they are losing $2bil every year. their last ditch effort to adopt linux just shows how much sun is failing.
you are stupid and don’t know how modern pcs work. the only way that could possibly stress a pc is if u were using one without 64bit pci, which would be really stupid. all the things you think are specific to sun computer are NOT.
You seem to be citing 64-bit PCI and SCSI as things I think are “specific to sun computer.” Certainly those technologies aren’t, but things like crossbars are. The bus architectures available on high-end Sun computers transcend most technologies available on PCs today. Perhaps we’ll see a shakeup with the Opteron and its NUMAlink architecture.
As for memory bandwidth, most manufactures of high-end systems solved this issue using parallel memory channels, similar to the nForce chipset on PCs. The UltraSPARC III supports up to 4 interleaved banks per processor, 2.4GB/s processor to memory transfers, and 4.8GB/s bus to memory transfers.
the numbers u give are impossible.
No, they’re not. Read them for yourself at http://www.sun.com/servers/highend/whitepapers/SUN_SF15K_DS_01-03_V…
the fastest fsb is the pentium 4’s and it runs at 533mhz
The Pentium 4 is a monument of inefficiency. Due to its extremely long 20-stage pipeline and relatively poor branch prediction, it is constantly mispredicting branches. Because it caches decoded micro-ops in its trace cache, a mispredicted branch renders its trace cache useless, so it must fall back on main RAM. Because of this, the Pentium 4 relies on excessive memory bandwidth to even work effectively.
u are saying u can read from disk faster than the fastest pcs can read from their ram? ha ha that is a major laugh buddy
No, but I am saying that Sun builds computers which can move much more data over their busses at a single time.
why ppl are so stupid and shell out thousands of dollars for a slow, expensive sparc computer when a pc can do EXACTLY THE SAME THING for a fraction of the cost is beyond me.
Unfortunately, this isn’t the case. Unless you can effect a load balanced decentralized solution, PCs aren’t an effective method of high performance computing. Any time data is centralized (for example, a central code repository or a database) a computer like a Sparc is needed.
people are just dumb and have loyalty to their sinking ships like sparc because they are so incredibly misinformed about pcs.
Or perhaps in the seemingly limited scope of your interaction with computers you’ve never seen a situation in which a computer like a Sun Fire 15k is warranted. Did that ever cross your mind?
In writing this reply, I’ve concocted a new theory. Perhaps the limited exposure that the IT world has to high-end systems is the reason that datacenters are replacing a single, expensive central server with a large farm of (initially) inexpensive PCs. Perhaps they’re simply uncomfortable with relying on a central server, and use a distributed solution to save their own skin…
This is probably not idea to ignite another flame in return but….
“ur sparc is using crappy anchient sdr memory…”
I really like the rabid use of abbreviations across your post, but the UltraSparc II sports a massive L2 cache ranging from 2 – 8 megabyte, or roughly 4x-16x what the current pentium sports.
“u can buy a motherboard with 64-bit pci…”
It may use 64-bit PCI slots, but thats like comparing a scanner and a dwarf star, they’re completely diffrent.
“ultra high speed ddr333 memory”
And you seem to be upgraded to To Niave 2.0 Super Ultra Turbo Pro! But, once again, the UltraSparc III has 2.4 GB/sec, while thats not quite the 5.3 GB/sec the DDR333 memory supports it can handle upto 16 Gigabytes per processor, and the 8 megabytes of cache adds up quickly (As shown in the jump up when the P4 went upto 512Kb).
“EXACTLY THE SAME THING…”
Not everything, if you have to deal with a massive matrix of values (a good way to calculate pi, mind you) the 16 Gigabytes of ram is priceless, because even the best PCs are limited to 4 Gigabytes and harddrive swapping makes the whole operation pointless.
And case in point”
“you are stupid…”, “why ppl are so stupid”… Shift key must be broken, but “EXACTLY THE SAME THING”. Scratch that.
Presentation is half the battle in the war for a clue, but Quake is a more fruitful pursuit for yourself it seems.
-If only you had a brain
Every heard of Cray? Well when they fell apart, most of the engineers went to two companies.. Sun and SGI. SGI pursued the NUMA arch and Sun pursued SMP high I/O bandwidth arch. around the UltraSparc. It’s not the fastest CPU, but that is not what makes a computer fast. Anyone who has experience with real HPC will know that is true. It’s how fast the PCI, Memory, Cache, Storage, etc. work together that make a system fast.
While you can play games on the PC and rave about how fast things are, you have to remember that it’s a game.. not a real-world computing job. Sun systems were designed to handle I/O intensive computing, which is why they are perfect for large databases like Oracle and SAP. The other big thing is that the UltraSparc is very scalable, as a matter of fact the UltraSparc line is the leader in near linear scalability. For everyday people, that may not be important. But let me put it into perspective.. if you buy software like Oracle or Veritas, they charge you a lot of
$$$ based on the number of systems. So buying tons of PC’s does not make a lot of sense in order to equal the power of a large SMP box. Any real business has some large SMP box running their business backends.
Sun’s computers were never meant to be used in everyday people’s homes for their desktops. That’s not Sun’s market, and quiet honestly they don’t care. Sun is in the business of making servers for customers who need a stable platform for running databases, simulations, websites, etc. The only reason that Sun is not doing well right now, is because of two things:
1. They are horrible at advertising.
2. They are a victim of their own success. There are more customers out there running Solaris 2.6 on 5 or 7 year old Sun hardware.. simply because it works and it works well.
Those are the real reasons for Sun not doing well. The used market for Sun equipment was about $9 Billion last year alone globally. And sadly, Sun didn’t make much of that money. That is why they pulled their agreement with GE Access to take back the certified used Sun equipment business.
When it comes to workstations, Sun’s market for workstations/desktops have shunken and caused the price and quality to drop to meet price points. Compairing a Sun Blade 150 to and Sun Blade 2000 is just wrong. The Sun Blade 2000 uses the UltraSparc III, FC-AL drives, dual-processor, better RAM, better backplane.. just better system all around. You could use the Sun Blade 2000 as a server or a high-end workstation. The Sun Blade 150 is more of a cheap desktop than anything else. As for the price.. the Sun Blade 150 is pretty cheap $999. The Sun Blade 2000 is a lot more expensive, but that is to be expected since it’s very simular to the 280R.
One of my biggest complaints about PC people is their completely backward thinking that clock speed and Mhz on a bus matters. Let me clue you in, it doesn’t. There isn’t a linear gain in performance by upping the Mhz on a PC. And that is undoubtably why AMD decided to not indicate the Mhz speed on their new processors. It’s a lot of hype. And to whine about Mhz on a BUS!! What a joke! The Sun Fire arch. is a cross-bar ECC protected switched fabric. And you may say, well what good is that. Well let me clue you in again.. the same arch is used for high end SAN’s like the Hitachi 9970 and 9980. Again, Sun is aiming for high I/O requirements as you can see by their designs. You just don’t get the same level of performance, reliability, or scalability on a PC, let alone a PC server.
The only reason that Sun has the LX50, x86 blade for the SB1600, and soon cheap PC running Linux for Sun rays is because there is a business market ready for such solutions. Companies are looking for cheap ways to reliably centralize applications and networks, and reduce desktop overhead. That is why Sun ported the Sun Ray client to Linux, it’s a cheap way to deliver an innovative product. If you have not used a SunRay, you don’t know what you’re talking about.
So you might want to take some time to educate yourself before putting your foot in your mouth. You only serve to demonstrate how stupid you are when you make such unimformed statements.
Another thing that makes a strong enterprise device is the manufacturer of the HW and writer of the SW under one roof… When one can control all the quality points of the system, only then does a extremely reliable and _generally_ fast system emerge.
Sun owns the HW and SW. Their chipsets are also made to be reliable under extreme load. Many PC chipsets still suffer from feature-outgrowth-itis, and have never been truly taken through a thorough shakedown to be sure there are no hidden states in the logic design.
I once wrote drivers for systems that used a famous Taiwanese chipset maker’s chipset (not VIA), and the chipset had state errors that could occur about 1 time every 7 days or so, and it corrupted RAM and locked the system. Only occured under heavy load — you would never see the bug otherwise. Turned out, it was a PCI bus “optimization” originally meant to speed _games_. So sure it sped up games (this was PCI bus at the time), but the chipset sucked for real world use. And you think this is only one example? Almost every chipset in existance for the x86 architecture that is currently available, has similar bugs, just waiting to attack your million dollar cashflow when they wedge a server… basically because of the very reason they are so cheap — standards are great, but they still have loopholes, and when you least expect it, someone will exploit those holes.
I use a fairly low-end Ultra 10 side-by-side with a PIII 933Mhz at work. While the Ultra10 is by no means a speed daemon, it’s never crashed on me like my Win2K PC, it’s uptime usually says when the last power outage occurred…
My company are close to being MS-only, but I think it says a lot that our application servers ALL run on Solaris/SPARC – and I’m talking 2000 fat clients for an application running the Oracle DB replication between the UK, California, and Singapore.
Low-end Suns may only have stability going for them, but high-end boxes are definitely worth their money is you value uptime and long-term investments.
Since there is so much talk about Sun hardware. How does it stack up to the Fujitsu PrimePower systems. http://primepower.fujitsu.com/en/index.html
PrimePower are based on Fujitsu’s SPARC64 and is ABI compatible with Sun’s UltraSPARC generation. According to benchmarks (SpecMark) I have seen, it oupaces UltraSparc significantly.
But I wonder how it fares in terms of I/O, Bus, Ram, reliability? Any comparisons?
It seems to me that Sun’s main competetive advantage is their system architecture (I/O) rather than their processor architecture. I wonder if it would be better if Sun phased out their own line of processors and just bought from other vendors. The best candidate would probably be future IA-64 (Itanium) chips (the only other candidate would be Power which is produced by Sun’s main competitor).
I know Itanium 1 sucked, but Itanium 2 is doing much better performance wise. My guess is that Sun could produce some pretty kick-butt Itanium 2 systems. The main problem I have with Sparc is NOT its performance, it is the cost to Sun. Can Sun afford to keep Sparc competetive? I honstly don’t know.
The worry is that Sun could lose its differentiating factor. They can’t compete on a commodity basis with Dell. Also, given HP’s close relationship with Intel on IA-64, would HP be able to produce better IA-64 systems or get better processor pricing from Intel.
Sun is in a tough spot. In the low end, they are being beaten up by Intel/Dell. In the high end, IBM is giving them pretty stiff competition. Their current strategy doesn’t seem like it is going to change things. AMD on the low end seems like a decent move (though Dell will still kill them on price). IBM will continue to compete well in the high end (with better processors).
Sun is in a tough spot. In the low end, they are being beaten up by Intel/Dell. In the high end, IBM is giving them pretty stiff competition. Their current strategy doesn’t seem like it is going to change things.
Sun’s current strategy is to employ massive parallelism in single die processors, see: http://www.aceshardware.com/read.jsp?id=55000245