Cell-based Server Blade Demonstrated

slashflood writes "Only a few clients in a hotel room near Los Angeles had the chance to see the first Cell based server blade running Linux 2.6.11. 'We demonstrated the prototype to show that Cell continues to mature. The product is expected to have several times higher performance compared to conventional servers,' said an IBM engineer."

Comments

I don't get it

[Anonymous Coward]

The Cell is just a PPC with 8 little miniprocessors tacked on. The miniprocessors have explicit control over and direct access to the contents of their own cache, but can only access data in awkward ways; and are super-optimized for vector/SIMD instructions and floating point operations, but are not so good at algorithmic or complex flow operations.

The Cell's bonus processors are absolutely great for DSP and multimedia apps, such as that we see in the Cell.

But, they are going to be at a strict disadvantage in data retrieval and pushing operations-- which is, incidentally, exactly what most servers, such as a file, web or database server, need to be best at!

What kind of servers *ARE* these??

Re:I don't get it

[chill]

Unless you buy into the whole "utility computing" paradigm, like IBM does. In that case, servers are going to be doing more than just handing up files and indexing databases.

Using a two-tier or three-tier approach to client/server architecture, with something like a full-duplex GigE connection to fat, diskless clients and you have some real potential.

A fat client (512+ Mb RAM, 1 CELL processor) that can use the backend for the more heavy-lifting tasks would be a fantastic setup for a lot of businesses.

-Charles

Re:I don't get it

[YU Nicks NE Way]

Sorry, but it really wouldn't. Think about it.

If you're running a rendering farm, then the Cell is a great tool. (And, if you think about it, a game console is essentially a low end rendering farm.) If you're running a word processor, however, SIMD instructions are useless. If you're performing a standard query against a database, SIMD instructions are useless. If you're sending an electronic mail, hey, guess what? SIMD instructions are useless.

I think that IBM Microelectronics is trying to Cell their new processor in the hopes of Celling their bosses on the (dubious) proposition that they can recoup the losses they've seen on their contract with Sony. They've packaged up the right buzzwords, and they're creating a lot of fog. I sort of doubt that it will work.

Re:I don't get it

[TheKidWho]

If you're running a word processor, a 3.2ghz processor is a bit useless too don't you think?

Re:I don't get it

[chill]

Yeah, but most "normal" apps like e-mail, word processing, web browsing, etc. are more than handled by the base processor. No, you don't get any help by the SIMD but you don't need it.

When you move to things like editing audio & video, print rasterizing, hi-res photo deforms and filters, then it'll kick in.

Think of what this sort of processing will do to GIMP/Photoshop filter speed... Or DVD/music ripping/encoding/decoding... Or audio mixing...

Re:I don't get it

[slashflood]

But, they are going to be at a strict disadvantage in data retrieval and pushing operations-- which is, incidentally, exactly what most servers, such as a file, web or database server, need to be best at! What kind of servers *ARE* these??

Servers you can find in the render farms of ILM. One of the demonstrations was a realtime ray tracing of a landscape. The resulting jpegs were streamed to an Apple G5, because the Cell-based blade server had no high end graphic board.

There are thousands of other applications for such a kind of server. The Earth Simulator is also not a file, web or database server.

On the other hand, even a web server can profit from a Cell server. Look at all the computations a PHP server is doing nowadays. A content management system like relies heavily on ImageMagick to generate the images on-the-fly. Look at all the content servers, like video and audio (mp3) download sites. Some of them are rendering thumbnails or converting uploaded content on the webserver itself.

Database servers are not only looking up entries in an index, they are also doing heavy calculations.

SIMD (like SSE) helps a lot in different areas. A file server could do real-time encryption...

correct me if i'm wrong....

[Lehk228]

Wast the benefit with Cell supposed to be that the programmable DSP's worked somewhat like pixel shaders except useful for all kinds of complex serial data so that operations on serial data could be massively improved, which does not seem to me like it would be a major help in a server, unless it is running a specialized app that just happens to be on a server for data access rather than using the Cell to speed up web servers etc.

Re:correct me if i'm wrong....

[NovaX]

I think its because web servers have thread pools, so a cell processor could handle many of these light-weight threads simultaniously. This makes it perfect for a blade server.

Sun's Niagara is aimed at this market, where the work is of great quanitity, not huge number crunching. This could mean searching, web page serving, and streaming media. So if you need to handle thousands of requests, this type of processor is ideal. Of course we won't truly know until one of these massively multicore beasts is out in the wild and can be tested in a realistic scenario.

! Graphics only

[theid0]

I've been trying to ignore everybody's outspoken assumptions about the Cell being a graphics chip which can't do general processing for a desktop computer. The fact is that it's rightly a multi-core chip with loads of vector processing capacity. It might not be as fast on a single-threaded task, but the software world is going to adapt quickly for this type of setup because it's where the hardware is going. No semiconductor lab can (cost) effectively compete in a megahertz race anymore, so more power = more transistors (more cores).

Server programs are ahead of the curve at this point because they've had multiple CPUs in abundance for a long time. However, even today it doesn't make sense for games like Doom III to avoid taking advantage of this hardware when possible (for instance, the G4/G5 systems have had dual processors for YEARS but Id won't use them properly). For petessake, calculate audio on one processor and AI on the other...

Re:I'm just curious

[BiAthlon]

Ok, I've got five mod points and you make me post something to this story instead of mod.

What would you do with a 'bit' that was "pretty close to 1" or "just a bit over 0"? You no longer have any exact state of data which every language I've ever used has depended on.

I like my 1's and 0's just fine thanks ;)

Re:OS X on Cell?

[nokiator]

There are several contexts which can lead to Cell processor being used in future Apple platforms:

• As a media co-processor in next generation PowerMacs, and potentially even high end iMacs, similar to "AV" badged Macs from a few years back. Cell can work as a pretty good general purpose media co-processor to offload video encode/decode operations from the main processor(s). Even the current high end dual processor PowerMacs are being challenged when decoding HD H.264. A co-processor that can enable real-time H.264 encoding would make a big impact on the user perception.
• As a physics modelling co-processor for Macs to accelerate animation and games. This is really what the Cell processor is designed for in the first place, and there is likely to be plenty of libraries/engines written for PS3. This will go a long way to eliminate the existing perception that Macs are inferior game machines. The same capabilities can be used by professionals for 3D animation work.
• As the core of a home media center that can encode/decode/store/stream video/audio. If the Cell can fit the thermal and cost constraints of a game console, it would also be a good fit for a next generation media center.

Re:Uhhh

[gorim]

The equivalent for Mac / PPC - altivec, velocity engine, or vmx (whatever you want to call it) certainly revolutionized that platform.

The fact that on the x86 platform there was little revolution, or one little seen, may be more a reflection of the platform itself.

Honestly, people who can't see the value of making true and powerful use of SIMD are missing the boat. That is what the future is all about.

You look at your cellphone, mp3 player, mp4 codecs, digital tvs and radios, it is SIMD that makes all that happen (through DSP).

More visible to you, look at your GPU's powering your favorite games, specialized SIMD.

The main CPU is truly just the conductor and SIMD is the orchestra, as Sony puts it.

Re:Now we just need to ask it tough questions!

[jericho4.0]

Comparing MFLOPS to intelligence comes from the same school that promised that true AI is "just around the corner" for 30 years.

The mind might indeed be a Turing machine, but it's a very different architecture and OS than the ones we know about.

probably only running on the central powercore

[nietsch]

Though it is very nice to see that IBM ported linux this quickly, I think they cut some corners. The cell has a central powerpc core, and 8 (or more) accesory processing units. The processing power lies in these APU's, not in the central power core. The APU are also very specialised, so you will ot only have to allow acces to the cell from the OS(and manage those), but you also have to write the userland programs that take advantage of the APU's strong points.
That applies to every program you want to use the apus, so the chance that this happens overnight/soon is pretty slim. Heck, they might even need to rewrite the benchmark programs for it.

Because they have not released any real benchmarks and only talk about theoretical numbers, i think they have not finished the porting fully (or have very disappointing benchmark numbers).

Giving early acces to LUGs would be nice for the street creds, but will not speed the code development of the mostly proprietary code that needs to run on it. Giving it to Gimp/Blender/other developers might work, if it comes with a crash course cell programming.

Re:show us the numbers

[Jules Labrie]

The Cell is basically a Vector Processor. So it's not gonna be really fast for compiling or such things like that. IBM just took the opportunity of the PS3 to develop the perfect processor for super-computing (whose task are often matrix-based). Server with cell ? No advantage, or so few. Games ? Becomes interesting, but that's all. Supercomputing ? Here you are.

I have to say, this Cell is really a great marketing coup ! Everyone is speaking of this processor, even in the biggest newspapers of the mainstream press... Not really seen that for a few years !

Re:IBM Blade Server Management

[turgid]

IBM has a vested interest in making things difficult and complicated for its customers. After all, it makes its money from support.