[computer-go] Pay-as-you-go cluster?

[Chrilly]

Yes MPI solves some of the problems. One gets at least back the info that 
something went wrong. Generally the best way do handle such a situation on a 
dedicated Cluster is to shutdown the programm.

Chrilly

----- Original Message ----- 
From: "David Doshay" <ddoshay at mac.com>
To: "computer-go" <computer-go at computer-go.org>
Sent: Monday, October 02, 2006 7:32 PM
Subject: Re: [computer-go] Pay-as-you-go cluster?


> This is true on @home and X-grid clusters, but is not a problem on 
> dedicated clusters. SlugGo uses blocking MPI calls to the child nodes 
> (the code waits for all distributed jobs to return before continuing)  and 
> has never hung waiting for a reply.
>
> Cheers,
> David
>
>
>
> On 2, Oct 2006, at 5:46 AM, Mark Boon wrote:
>
>>
>> On 2-okt-06, at 09:35, Darren Cook wrote:
>>
>>> *:asynchronously: once all leaf nodes have replied the minimax  tree is 
>>> made.
>>
>> Using a cluster is only going to work reliably if you design for  the 
>> possibility that a node will never reply. One way to do that is  assign 
>> nodes that finished already to the same task(s) of nodes  that didn't 
>> reply yet.
>>
>> Mark
>>
>>
>> _______________________________________________
>> computer-go mailing list
>> computer-go at computer-go.org
>> http://www.computer-go.org/mailman/listinfo/computer-go/
>
> _______________________________________________
> computer-go mailing list
> computer-go at computer-go.org
> http://www.computer-go.org/mailman/listinfo/computer-go/