GCG server

Rick Westerman westerman at purdue.edu
Wed Oct 9 11:18:56 EST 2002

nelly wrote:

>Really thanks for your quick answer. If you could just tell me a little more
>about the number of processors of the server if I consider ther would be 40
>active users ( the others working in batch): I think  4 would be good but I
>m not sure. and about the E-cache, 6 GB (or less)?? I think that it would
>have a need of about 0.5TB of internal/external storage for
    That is about right.

>I d like to know if you would choose an entry server or a midframe
>server(SGI300, Sun enterprise4800)....
    It really depends on your finances.  The entry level servers will be 
cheaper per CPU because they do not have the extra hardware required by 
the midframe machines.  Midframe machines may have a different (and more 
expensive) pricing structure than the entry level servers.  As an 
example, when we were purchasing a GCG machine several years ago we went 
with a Sun E450 (4-CPU max, 4-GB memory max, lots of disk capacity) over 
a Sun E3500 (8 CPU max).  Why? Because we could purchase a fully loaded 
4-CPU/4-GB E450 for the same price as a 2-CPU E3500.   Since no 
additional funds were looming we decided to go for the 
limited-expansion-but-many-CPUs machine; i.e., the E450 instead of the 
E3500.  Of course 8 months later when more money was shaken free and we 
we able to purchase more hardware I started thinking dark thoughts about 
the upper administration.  We ended up with three maxed-out E450s 
instead of one maxed-out E3500 -- which we could have had given proper 
financial forecasting.  The three machines allows us to differentiate 
into user types -- GCG, Statistics, Database -- but does not allow us to 
use the processors efficiently.

   To forestall the next question:  yes, our 60+ users, run GCG on one 
E450 just fine.  However when we do a large scale blast search (100s of 
sequences) we offload this to other machines.  Think Linux cluster for 
this type of work.

-- Rick

More information about the Info-gcg mailing list

Send comments to us at biosci-help [At] net.bio.net