Re: Grids and Clouds

Greg Pfister wrote:

On Jun 15, 10:58 am, Geoffrey Fox mailto:g...@grids.ucs.indiana.edu wrote:   

There are 100's of Grid projects especially in universities and government arena. They have quite complicated software stacks such as   Glite UNICORE Globus Open Science Grid, OGSA etc. There are discussions of linkage of Grids and Clouds includinghttp://www.ogf.org/OGF23/materials/1303/Grid+cloud+comparative+study+...     

 Thanks for the link. However, it seems to mainly be a "Why is EGEE better than Amazon EC2" pitch. 

Interesting you read this into talk. The speaker at meeting came over IMHO as favoring clouds v. EGEE. The claim from EGEE in audience was that authorization system (VOMS) in EGEE could not be replicated by EC2. This was not to me a very convincing issue (maybe EGEE doesn't need this complexity. Probably could build VOMS as an addon to EC2 if needed to)

But it does raise a question for me:  Are the capabilities portrayed for EGEE typical of Grids? It looks like it's compute and storage allocation, workload management, file / filesystem management, user/group access & control.   

I would consider EGEE as perhaps the most sophisticated "compute-file" grid. The Open Science Grid is USA activity with similar capabilities and software. TeraGrid and DEISA (European collection of supercomputers) support similar computing model but with fewer much larger base systems.
There are class of more data (database, sensor) oriented grids which are quite successful in areas like astronomy/bioinformatics/environmental science etc. These are built around the intrinsically distributed nature of data in many applications.

 If that's basically it, then lots of clouds provide lots of commercially-oriented middleware that doesn't appear in Grids. This would make sense, given the origins of both notions.  Again if the above is true, the answers to the questions below follow -- (inserted below)    

and some growing expectation that cloud ideas/technologies should/will be used in Grids. One key question could be  a) Can Grids evolve usefully to be or to include Clouds. If so, what is evolution path?     

 Sure, add all the commercial system support: databases, firewalls, transaction orientation (throughput orientation) providing simple scaling, etc.    

b) Are clouds just different and one needs to start again possibly building selected Grid capabilities (such as Condor discussed often in this list or security infrastructure like VOMS) on a cloud platform/infrastructure?     

 I suspect the biggest difference is the throughput orientation of many clouds. If your workload is lots of little requests,

Sensor Grids would have lots of little requests

 it's a fundamentally different game from one humongous galactic cluster simulation. -- Greg Pfister       

--  : : Geoffrey Fox  gcf@indiana.edu FAX 8128567972 http://www.infomall.org : Phones Cell 812-219-4643 Home 8123239196 Lab 8128567977 : SkypeIn 812-669-0772 with voicemail, International cell 8123910207

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups "Cloud Computing" group.
To post to this group, send email to cloud-computing@googlegroups.com
To unsubscribe from this group, send email to cloud-computing-unsubscribe@googlegroups.com
For more options, visit this group at http://groups.google.ca/group/cloud-computing?hl=en
-~----------~----~----~----~------~----~------~--~---