IBM Touts Supercomputers for Enterprise

Stony Stevenson writes "IBM has announced an initiative to offer smaller versions of its high-performance computers to enterprise customers. The first new machine is a QS22 BladeCenter server powered by a Cell processor. Developed to power gaming systems, the Cell chip has also garnered interest from the supercomputing community owing to its ability to handle large amounts of floating point calculations. IBM hopes that the chips, which currently power climate modeling and other traditional supercomputing tasks, will also appeal to customers ranging from financial analysis firms to animation studios."

Oblig..

But is it Vista capable?

Re:

What am I doing wrong because, hard as I try, I am unable to get my Vista box to crash. Please let me know what I have to do to get it to crash.

Java? Who cares?

Completely wrong. This is industrial supercomputing, not random web applications. The applications IBM is targetting Cell at are things like seismic analysis for oil companies, and this requires highly tuned implementations and specialised algorithms. This is expensive. It takes a lot of developer hours, and those developers are top-rung, very highly paid, because this stuff is very hard. And it costs a lot in hardware. Unfortunately for IBM, while Cell is fast, it's not fast enough to justify the cost for most companies. And it has a lot of competition from NVIDIA's Tesla platform, AMD/ATI's FireStream, and plain old clusters.

Re:

It's only completely wrong if we assume the story is a non-story. The markets you're talking about are traditional super-computer markets. The summary at least (too tired to read the full story today) is suggesting that IBM is targeting "Enterprise" custom

Re:

Completely wrong. This is industrial supercomputing, not random web applications.

Unless they know something we don't... Could it be that Duke Nukem Forever is about to be released ?

The trend towards commodity hardware continues...

I've seen more than my share of traditionally big iron applicatoins (databases, data warehousing etc..) being moved off of specialized hardware (ibm p595s, sun e15k, HPSuperdomes etc..) being moved onto (or attempted to) commodity hardware. Management hears,

"We can replace that $2m server with 10racks of servers each $1k and after 3yrs we just throw them away and replace them with the latest and greatest x86 based hardware with 2x performance still $1k/server?

Now IBM wants to push highly specialized blades. Somewhere someone's saying, "How many x86 servers can we get for one Cell blade?"

Personally, I'm sick of managing farms of physical servers, and with the introduction of VMWare, I'm now managing 3x the number of machines (albeit virtual machines). Have an FTP server? Run that in it's own image. Also have a syslog server? Yet another virtual machine. I really hope this sells well. Maybe I can now play PS3 games in the datacenter.

Re:The trend towards commodity hardware continues.

If specialized hardware returns to vogue, then there problems will crop up with the new specialized hardware. Dude face it, if you are a sysadmin, God will provide you with your share of things to complain about. It is the natural working order of things.

Re:The trend towards commodity hardware continues.

> Personally, I'm sick of managing farms of
> physical servers, and with the introduction of
> VMWare, I'm now managing 3x the number of
> machines (albeit virtual machines).

All those Virtual machines to do the same thing with 4 times the resources as one well configured Linux box. Tsk Tsk.

Oh, but don't you LOOK busy.....

Re:The trend towards commodity hardware continues.

> Have an FTP server? Run that in it's own image. Also have a syslog server? Yet another virtual machine.

Which as you have no doubt discovered, are sort of a PITA to administer because they're all in separate VMs. I suspect the next big thing in commodity server virtualization will be nice management interfaces and protocols that break down some of the management walls between VMs, while still leaving the more important parts of the virtual environment intact. And being able to change the hardware assigned to a VM on the fly will probably become more common, too. I'd give it 5 or 6 more years, and VMWare will probably have managed to reinvent the LPAR.

Gotta love how this stuff goes around in cycles. Anything cool today in microcomputers was probably boring people to tears 10 or 15 years ago on large systems. (Cf. multitasking, multiple users, parallel processing, network-oriented filesystems, virtualization, hypervisors...)

Re:

Totally agree on virtualization...over-hyped BS.

Don't get me wrong, it definately had its uses and we do use it some where I work, but it's not the be-all end-all that VMWare and co. would like you to believe. Basically if you're not using it for:

a) Test

Re:The trend towards commodity hardware continues.

I've seen more than my share of traditionally big iron applicatoins (databases, data warehousing etc..) being moved off of specialized hardware (ibm p595s, sun e15k, HPSuperdomes etc..) being moved onto (or attempted to) commodity hardware.

I'm honestly curious. How well does this really work out for databases and data warehousing?

One of the benefits I understand from going with one big-ass server is that the memory to pipelines between cpu and memory are much better than the ethernet/myrin

Flamage

Developed to power gaming systems, the Cell chip...

That's just what we need to introduce to enterprise computing, the flame wars and invectives of the console world.

Admin #1 cell totally rockzorz!!!11!!!1 u n00bs using virtualization are in the past
Admin #2 IBM SUCKS!!! YOU KNOW IT!!! YOUR WHOLE BUSINESS IS TOTALLY GAY FOR THEM!!!!

Re:

Actually, the Cell processor is a rocking piece of hardware. I'd like to see something like it added to motherboards as an optional coprocessor arrangement. Yes, I realize that the code/compilers would have to be redone, but I think that this is something

Re: Flamage du jour?

A coprocessor for what? You do realize that each Cell already has 8 SPEs attached to it, and it's designed to be tightly paired with another Cell? It already delivers kind of ridiculous bang for the buck, so I'm curious what you're asking for.

(Usual discla

Re: Flamage du jour?

Even though this might make me sound a bit off, we already have co-processors for video, network, etc. Why not go a bit further and specialize the hardware just a bit more. Let the Cell do all the real work and sandbox the user on the x86 cpu in a way that allows the user to be rather free in operation while the real work is done on the Cell processor in protected manner. That "should" be enough processing power to isolate the user completely from the tasks of the computer itself. The idea would be to sort-of create a mainframe/client environment where it would be nearly impossible for the user to accidentally introduce viri to the system.

The UI and i/o shelled through the x86 system. There are examples of this in some smaller embedded systems where system memory is separate from user memory etc. The details of this seem sketchy as I have not worked them out to any degree that would make the proposal sound workable thus far. I do know of examples where techniques like this are used to protect the 'system' while 'user agents' do what they want without the intrusion of security software at every turn. When the system is turned off, the user space is cleared. The protected system space is always protected.

Yes, that leaves room for infections on the Cell side to act like root kits as there is always some spot that is vulnerable, but it does offer a much more bullet resistant setup. The effects are not too different from working from a live CD all the time. Reboot and all is clean again, but with a more permanent and less inconvenient process. If you run some version of Linux/Unix on the client side, and strictly control the communications to the Cell side it becomes a much tighter box to try to squeeze a virus into. It may provide opportunity for the Cell side to monitor processes in the client/UI side meaning that keyloggers and such wou9ld become a thing of the past. In general, I mean to add horsepower by splitting system tasks from UI tasks and add a much stronger sandbox for the client to operate in, rather than continue lumping all the work on one cpu and letting security run in the same sandbox as the questionable software.

It's an idea... obviously I do not design motherboards or OSes for a living (IANACSPHD ??)

Re:

Sorry, but I think you're off on a pretty wild tangent as to what the Cell processor is really about... However, I don't feel that I'm in a position to do more than to refer you to the publicly available documentation. There's quite a bit available about t

Let me guess

LCARS

no I'm not yelling, you dillhole

Re:Let me guess

Heh. Like with you, the first thing that popped into my head upon reading the headline was, "Wait, IBM make supercomputers for starships now? Since when?"

Scotty was heard to say...

SCOTTY: Computer. Computer! Hello computer!

SAM PALMISANO: Uh, just use the keyboard.

SCOTTY: Keyboard... how quaint.

Yes, It Does Run Linux

From IBM's detailed press release [ibm.com]:

the QS22 boasts an open environment, utilizing the flexibility of Red Hat Enterprise Linux as the primary operating system and the open development environment of Eclipse.

That means that a PS3 running Linux [psubuntu.com], even with its ridiculously low 512MB RAM, can be used as a $500 development platform for these CellBE BladeServers.

And, in turn, some QS22 SW might be usable on the PS3, if it can be ported to use the tiny RAM. Or if someone hooks an i-RAM bank to the SATA port as swap/ramdisk, using perhaps iSCSI over its Gb-e for storage.

This is a Failure

First off, I should admit that the Cell is quite a good processor, however it has shortcomings. Many shortcomings.

I'll begin first by listing the positives:
It has very good single precision floating point performance
Very high bandwidth on chip (25 GB/s)
With the e[nhanced]D[ouble]P[recision] addition to it IBM is adding for Los Alamos' Road Runner, it should be even better.
IBM Developerworks is a useful resource for programmers
It's a processor that gives the academic community a chance to publish more papers

The negatives:
Far too expensive
Floating point performance, single or otherwise, is useless for most enterprise work (no point running a database on a Cell)
The things that makes the Cell fast and unique are the SPEs (SIMD processors) are useless for servers*(I'll elaborate below)
Developing software to make use of the SPEs is time consuming and difficult. Orders of magnitude more so if your algorithm isn't suited for being split across SPEs
Computer Science or Engineering students who are doing research in this field have to read redundant papers that reword the IBM Cell manual (plagarism?) for 70% of it and what they actually did with the Cell for about 10%. 20% is of course left for citations.
You can't actually play games on a Cell. The PS3 games use them for physics engines, sound or such; the video is done on a video card.
Worst of all, YOU HAVE TO GET DATA FROM MAIN MEMORY TO SPU'S CACHE YOURSELF! (also known as Local Store, 256KB)

Now to talk about the SPEs... They are what makes the Cell tick. If the Cell didn't have them, it would just be an old Mac processor. (IBM Power 4 was it? I've trying to avoid research papers on the Cell) The graduate students (a year away from a PhD) that I was observing who were doing developing a scientific application on the Cell.. when we summed their year in development, they essentially told me it was a pain in the ass. Why? Because you have to get data from main memory to local store. Imagine if you had to get data from memory to your x86 processors L2 or L1 cache. (No, you don't actually want that. Trust me, you don't.) Scientists don't want to develop applications on this platform. Researchers like it because they can publish papers talking about it. This is why Los Alamos' Road Runner might be a flop. Bye bye to $100m of US tax payer money.

Back to the discussion with the grad students... I then came up with the idea that the Cell would be perfectly suited to an Asterix server. Why? Because sometimes Asterix needs to convert from one format to another when audio codecs differ. This is a perfect SIMD application, except we have a problem. While the Cell does have great on chip bandwidth, it has very poor Power5 processor (PPU) to SPU latency*. There is a research paper out there that puts PPU initiated memory transfers to SPU latency at about 4 microseconds. Compare this to SPU initiated transfer to SPU latency of about a tenth of a microsecond. Huge difference. You can't really avoid this unless you want to develop something horrendous that avoid the PPU but that's unlikely. The PPU would run the actual server and then it would need to notify the SPU of data, either by sending message (slow) and the SPU getting the data itself (2*slow=2slow) or sending the data and a message (slower).

It will be about 5 years until good software development tools are written for this architecture that will address most of the issues. This is assuming that IBM doesn't ditch it, which seemed likely before this announcement.

Oh, by the way, the QS22 doesn't support a hard drive which means you need fast NFS. The cost of ownership is more than just the cost of single blade.

Re:

Your post could have been much shorter if you wrote it this way: "This will fail because there are tradeoffs for using it. I know this because I'm in college and other college kids said it was hard to use. I tried to come up with an idea, but there ar

Re:

There's only one Pixar and there's only one ILM. The other million animation studios out there don't have budgets even close to these guys, particularly considering the turn around on hardware (today's super cluster is tomorrow's pile of junk). Renderfarms will be staying with cheaper vendors (which also means white box for most) for some time to come yet.

I understand your argument, but there's an upside. Maybe someone will put energy into Cell processor renderfarm software, so all the fiscally-challenged shops can buy a small rack of PS3s and go at it? IBM or not, the Cell is still a damn quick CPU for se

Re:Which Enterprise?

N-C-C one seven oh one. no bloody A, B, C, or D.

Re:Which Enterprise?

You make it sound like it's a bad thing...