Slashdot Mirror

If you are interested in processing power and parallel applications, check this out: Mathstar. Their single chip, operating at 1GHz can churn out 500 billion ops a second. Just think that 4sq. ft. board is now on one chip - and much cheaper. Just think if you could wire 2K of these together (they are like $250 or something a piece in 1K lots) - we are looking at another petaflop contender...

Here is one answer: The FPOA

Depending on the nature of your workload, a 400 object Field Programmable Object Array (object ~ core) might be a better choice than a tradition CPU.

Check out the company Mathstar (http://www.mathstar.com/). They just taped out a chip the other day that when it comes to market will do about 500 Gflops a chip. The technology is quite incredable and although it is not specifically a general purpose chip the chip can be programed to work in any way you like allowing you to get max preformance for the applications that you need to run. Honestly I would like to get a hold of about, say, 50 of these and see what I could make them do in parallel (as they are made to be hooked up in parallel also). From what I have heard these would be competitive with processors now a days in price and therefore likely less then $1,000 a piece. Making it $0.50 a Gflop!

Sounds like you're describing the ultimate reconfigurable computer. Something that can , on the fly, reconfigure itself and change its architecture to the needs of the user. You're right, this would be great. Massively parallel and reconfigurable chips DO exist yet they are far from mature (http://www.mathstar.com./ If you're suggesting that the architecture should just magically morph from parallel to more parallel..I think you're asking a bit too much.

The cell chips are interesting, but from the description they seem to be more of a variant of the current multi-core chips that Intel, Sun, HP, IBM, et. al, are doing. It sounds like they have the capability of putting more than one type of CPU on the same silicon, which is different, but it is still a variant of existing designs. It doesn't sound like they will be putting more than a hand full of cores on their chips. Sun is already talking about 8 cores on their new Throughput Computing chip line now, and more in the future. Sun claims that their Throughput computing chips will ultimately be 30x more powerful than what is out there today.

If you want to look at something really different, check out the Mathstar FPOA chips. Right now they can put up to 400 1Ghz processors / devices in different mixes on a single programmable chip. There will probably be a lot of applications that this will be a better match for than a multi-core IBM Power based chip, although in fairness they are probably targeted at different applications. On the other hand, maybe the FPOA will be the "IBM PC" of the multicore chip world. When the IBM PC first came out it was laughed at by the IT shops as being so small and limited, that it wasn't a "real computer". But the people who had PCs didn't have to wait weeks for the IT Department to rewrite their program, run their report, or crunch numbers for them on the mainframe. The PC changed the world. I wonder what the FPOA will do?

What you're asking for is a SAN.

I just installed a Network Appliance FAS250 in my server room. It speaks CIFS, NFS, and iSCSI.

By the way, you're wrong... Oracle will run perfectly using CIFS shares (I'm running it now, and have been for the past few months), and NetApp has plenty of documents in their tech library showing all the different ways to use attached storage with Oracle and many other pieces of software.

With respect to speed, it really depends on the network infrastructure. I've got a Cisco GigE switch attaching 6 machines directly to a GigE port on the NetApp Filer. It is literally twice as fast than the directly attached RAID 5 (caching, etc.) arrays that it replaced.

I think that Microsoft Exchange can be installed to a CIFS share, but if not, you should look at iSCSI. My company uses Lotus Notes 4.6.7 (sweet, merciful Christ, please put me out of my misery), and it works great from a CIFS share on the NetApp.

Microsoft has a free iSCSI Initiator for Windows that will mount an iSCSI device just like any other SCSI drive in Windows. You can find several iSCSI targets for linux here.

I have about 50 Mac's on our network (graphics department) that needed to talk with the new filer. Instead of installing a klugy piece of software to make the OS9 Macs talk to the SAN at $150/seat, I installed a linux box using samba to talk to the SAN through CIFS and netatalk (AppleTalk for linux) to re-share out the samba mounts. Becides some quirks (Mac's don't see the linux gateway in the AFP browse list, but can connect directly through IP), it works rather well.

Look at iSCSI, it does exactly what you're looking for.

Building multiprocessor chips, or chips from arrays of processors has become a fairly hot design approach. There are a number of companies using it. It seems to be especially popular in the reconfigurable computing area. There is an interesting paper here. These processors go well beyond the current crop of dual CPU core chips like the P4, Power 5, and Ultrasparc IV.

Clearspeed's chip is a static 64 processor array chip aimed at FPU intensive applications, but there are many more things that you can do with array designs.

Mathstar is building a reconfigurable chip with hundreds of elements availble in various mixes of processors, memory blocks and other components. They are trying to replace ASICs and FPGAs as a platform for some part needs. There was a story on their architecture in EE Times a couple of months ago.

Intel is wokring on an array based processor aimed at the radio / communications market. I will be interested to see if their work with these chips ends up being used in other Intel chips. That could be deadly. So, the Pentium-X sucks at that task today? [Morph] Not now!

Phillips has what they call Silicon Hive technology which is another reconfigurable processor of functional blocks.

There have been plenty of companies using arrays and reconfigurable techniques too, like Altera and Chameleon.

Sun bought up a start up and is developing massivly multithreaded processors based on the start-up's technology. They call it Throughput Computing. They claim that in about two years they will have a chip 30x faster than todays designs. I'll be very interested to see if they can do that.

The next couple of years should be very interesting on the processor front.