[Bioclusters] FPGA in bioinformatics clusters (again?)

[Tirath Ramdas]

Hi Joe and all,

Please excuse the ridiculous latency of this response... been on  
holiday :)

On 13/02/2006, at 4:33 PM, Joe Landman wrote:
> I/we strongly support their use.  Not as replacements for clusters,  
> but as tools to significantly augment desktop and cluster level  
> supercomputing in Life sciences and related fields.

Noted, and agreed.


> The issue is that you have some design work to do to interface the  
> core to the rest of the board.  Now if the boards were somehow  
> standardized (cough cough) this would be a "good thing".  [anyone  
> want to talk about standard boards?]

I see your point. There's some irony here... the problem is in ``glue  
logic''. That's exactly what FPGAs were considered good for in the  
early days! But yes, I appreciate the significance of the integration  
headache, especially considering where we want to take this: i.e.  
making FPGAs available to end users. You might want to pop by over to  
the OpenFPGA list; I recall seeing some discussion of integration  
related matters, even to the Operating System level, though to be  
frank right now that isn't really a core technical interest of mine.


> I would argue the opposite, that bioclusters is all about providing  
> scalable platforms for bio-computing tasks, and that acceleration  
> systems, as people need them require a platform to host them.  What  
> better platform for a bio-accelerator than a bio-cluster ? (note:

Yep, fair enough. My concern was that we were getting a bit away from  
the bio side of things, but you pulled it back, so it's all good -  
believe me, I'm quite happy to go on and on about FPGAs and  
application specific processors :)

Actually, I have a question for folks who actually have access to  
FPGA equipped systems (I have a bunch of FPGAs lying around, but the  
very BEST system comms available is USB - and there's a Spartan  
sitting on that one!!); sorry for dragging vendor names into this,  
but I think it's probably the best way to illustrate the nature of  
the system I have in mind - I'm speaking of systems of the ilk of the  
Cray XD-1, Altix, and SRC's boxen. How much of a "multiple-use"  
attitude is there with these systems?

Bioinformatics is great, but other tasks - e.g. network intrusion  
detection, molecular dynamics, dense linear algebra - are also at  
home on FPGAs. The issue is multiplexing of the resources for shared  
access in a multi-user system, and this actually builds on top of the  
need for standardized integration, which presumably is already  
somewhat sorted on the systems mentioned. In a real world  
"production" environment, what actually happens?