[Bioclusters] General question on time consuming problems

[Tim Cutts]

On 20 Apr 2005, at 2:25 pm, Joe Landman wrote:

> Hi folks:
>
>  Are computational bottlenecks the major problem you are running into 
> today?   What do you see in the future in terms of rate limiting 
> efforts?  If you had an "infinitely fast" cluster (like a blue-gene 
> from IBM), how would like impact your work/processes?

The major bottlenecks are based around IO.  We have plenty of CPU grunt.

Scalable databases and parallel filesystems are what we need to sort 
out now.  It's no use having infinite amounts of CPU power if you have 
to force all the output through a very tiny pipe.

A lot of this can be solved by programming expertise, but most 
scientists aren't interested in coding for scalability, they're only 
interested in quickly producing something which produces "the right 
answer", whatever that means.

Having scalable filesystems and databases would allow them to carry on 
coding in their current less-than-perfect ways and still maintain some 
half-decent performance.

Tim

-- 
Dr Tim Cutts
Informatics Systems Group, Wellcome Trust Sanger Institute
GPG: 1024D/E3134233 FE3D 6C73 BBD6 726A A3F5  860B 3CDD 3F56 E313 4233