[Bioclusters] Distributed Load Management on the desktop

[Tim Cutts]

On 10 Apr 2006, at 7:44 pm, Andrew D. Fant wrote:

>
> This does beg a question.  Do your users do all their GUI-work and  
> Visualization
> over the network and run those software packages on the head node?   
> I'd love to
> have the cluster with isolated storage and such, but I've been  
> trumped by
> administrative policy.

There isn't a great deal of graphical work going on, to be honest.   
Those people that are running GUI applications (sequencing finishers  
running ACeDB, gap4, dotter etc, and microarray people running  
BioConductor stuff) are generally running them on tight-ish coupled  
Tru64 clusters using a cluster filesystem, and displaying to their  
desktop X server, so yes, they are running on the head node, but the  
head node is usually actually pretty meaty, at least as far as its I/ 
O capabilities are concerned.

We're not planning to change this idea radically, but building  
similar architectures out of open source components is an interesting  
exercise.  There are various bits and pieces which can provide much  
of what TruCluster does on Tru64 (heartbeat instead of CAA, and  
Lustre or GPFS instead of AdvFS/CFS, for example) but they're not as  
mature.

Tim