[prev in list] [next in list] [prev in thread] [next in thread] 

List:       npaci-rocks-discussion
Subject:    Re: [Rocks-Discuss]cluster planning questions
From:       Laurence Liew <laurenceliew () yahoo ! com ! sg>
Date:       2003-11-24 12:43:30
Message-ID: 1069677808.2179.81.camel () scalable
[Download RAW message or body]

Hi,

I would be interested to see the architecture for the below....
questions

> We just helped a group spec a 128 node cluster (2 head nodes, 9 NFS servers, and
> 128 compute
> nodes). Only the compute nodes will have myrinet -- everything else travels over
> gigabit ethernet.


1) 2 head nodes => High Availability?, rsync for mysql files?

2) 9 NFS servers -> how do you distribute the read/write to the 9 NFS
file servers? - via a load balancer or just 1 NFS to 16 nodes and do a
rsync of the NFS servers

Maybe on the rocks website we should have a best practise of
8,16,32,64,128, 256, 512 design.. sort of a "cluster pattern" reference
for both experienced and begining cluster builders.

We could base it just like the "gang of four" book on Patterns in
software engineering....

we capture the best practises and plus and minus of various cluster
designs...


CHeers!
laurence


[prev in list] [next in list] [prev in thread] [next in thread]