Characterizing computing capacity in heterogeneous environment - the Amazon approach
Dear all, I suggest to give a look at the Amazon approach about the characterization of computing capacity that is provided in heterogeneous an on-demand environment: http://www.amazon.com/b/ref=sc_fe_c_0_201590011_2?ie=UTF8&node=370375011&no=201590011&me=A36L942TSJ2AJA Cheers, Sergio -- Sergio Andreozzi INFN-CNAF, Tel: +39 051 609 2860 Viale Berti Pichat, 6/2 Fax: +39 051 609 2746 40126 Bologna (Italy) Web: http://www.cnaf.infn.it/~andreozzi
hi Sergio, I really like the concept of "Compute Unit". It is both descriptive and ambiguous. :) However, they don't address the shared memory problem. I also like the comment "to find out which instance will work best for your application, the best thing to do is to launch an instance and benchmark your own application." This suggests that we should give brief descriptions and if you really want to know details you should submit a "probe". I think that this is very important to remember, we are describing the resource but don't guarantee what is there and should give only high level descriptions to "steer" workflows. If you need to know something important which is quite specific to your application, then you should always run a test job to get the information that is important to you. Laurence Sergio Andreozzi wrote:
Dear all,
I suggest to give a look at the Amazon approach about the characterization of computing capacity that is provided in heterogeneous an on-demand environment:
Cheers, Sergio
Pretty much what LHCb pilot jobs do.
-----Original Message----- From: glue-wg-bounces@ogf.org [mailto:glue-wg-bounces@ogf.org] On Behalf Of Laurence Sent: 30 January 2008 14:50 To: Sergio Andreozzi Cc: glue-wg@ogf.org Subject: Re: [glue-wg] Characterizing computing capacity in heterogeneous environment - the Amazon approach
hi Sergio,
I really like the concept of "Compute Unit". It is both descriptive and ambiguous. :) However, they don't address the shared memory problem.
I also like the comment "to find out which instance will work best for your application, the best thing to do is to launch an instance and benchmark your own application." This suggests that we should give brief descriptions and if you really want to know details you should submit a "probe". I think that this is very important to remember, we are describing the resource but don't guarantee what is there and should give only high level descriptions to "steer" workflows. If you need to know something important which is quite specific to your application, then you should always run a test job to get the information that is important to you.
Laurence
Sergio Andreozzi wrote:
Dear all,
I suggest to give a look at the Amazon approach about the characterization of computing capacity that is provided in heterogeneous an on-demand environment:
http://www.amazon.com/b/ref=sc_fe_c_0_201590011_2?ie=UTF8&node =370375011&no=201590011&me=A36L942TSJ2AJA
Cheers, Sergio
_______________________________________________ glue-wg mailing list glue-wg@ogf.org http://www.ogf.org/mailman/listinfo/glue-wg
participants (3)
-
Gordon, JC (John) -
Laurence -
Sergio Andreozzi