[Bioclusters] Chargeback(s) for cluster use?

Chris Dagdigian dag at sonsorol.org
Wed Jan 5 14:37:11 EST 2005


I'll second Andy's words.

I rarely (if ever) see real world chargeback systems on lifescience 
clusters.

Most people I know use the chargeback hooks and accounting files to 
track resource consumption by user/application/department/project simply 
so they have a good idea what is going on and can generate nice reports 
for management.

Far more common is for departments, users or projects to pay for compute 
nodes or pay to add additional compute nodes to the cluster. When the 
cluster is idle, anyone can use any node. When there is contention the 
scheduler allocates resources (usually) by a percentage basis so that 
the larger "owners" get their faireshare of the contested resources. 
Both LSF and Grid Engine have scheduling policy mechanisms that make 
this really easy to do.

-Chris




Andrew D. Fant wrote:

> Kent Brodie wrote:
> 
>> Greetings - New Bioinformatics forum user here.   I expect to be an 
>> active
>> participant :-)
>>
>> I've been asked to look into other organizations that have a 
>> 'bioinformatics
>> cluster' similar to ours, and to find out if any of you have any sort of
>> "chargeback" scheme in place for users?  (By users, that could be other
>> departments within your organization, or even outside entities?).
> 
> 
> Kent,
>     Welcome to the list.  My best advice about chargeback is to avoid it 
> if you can.  I've seen a couple of good sites killed by a move from 
> funny-money allocations to hard currency chargeback.  Even if people are 
> willing and able to pay hard currency per cycle/byte/block for the 
> resource, you very quickly run to a point where it's almost a full-time 
> bookkeeping job just tracking all the costs of running the service and 
> apportioning them appropriately.
> 
> The ideal way to fund shared resource bioinformatics clusters is, of 
> course, to be funded as an overhead shared resource.8-).  Failing that, 
> my approach would be to go to the groups/departments that would 
> utilizing the resource and ask them to pony up a share of the costs on 
> an regular basis, and use a fairshare scheduler on the cluster to make 
> sure that the groups get cycles in proportion to their support.   If 
> your user communities trust each other, a simple model of paying up 
> something for access is also easier, but be sure you don't get into the 
> position of being personally expected to arbite disputes between users 
> about resource hogging.
> 
> Hope This Helps,
>         Andy
> _______________________________________________
> Bioclusters maillist  -  Bioclusters at bioinformatics.org
> https://bioinformatics.org/mailman/listinfo/bioclusters

-- 
Chris Dagdigian, <dag at sonsorol.org>
BioTeam  - Independent life science IT & informatics consulting
Office: 617-665-6088, Mobile: 617-877-5498, Fax: 425-699-0193
PGP KeyID: 83D4310E iChat/AIM: bioteamdag  Web: http://bioteam.net


More information about the Bioclusters mailing list