[gridcpr-wg] Use Cases for Grid Checkpoint and Recovery - Comments

[guru prasad]

Hello,
 
Following are some of my comments / concerns with respect to "Use-Cases for Grid Checkpoint and Recovery" document open for public comments. Please note that this is the first time I am submitting comments, if I have violated any etiquettes in my writing please excuse me.
 
1) Figure 1 and lines 253 - 255

The first thing to notice is that the box labelled “Application” contains a box labelled “Computation”. Some of the use-cases envisions an existing computation being directly modified to interact with a GridCPR system.

 

This places an assumption on the application to have "computation" as a part of the application. But when I think of in general and the applied fields of Grid Computing, not all the applications have computing as a part of it. (Eg. An application to collect all the x-ray of a user from a remote server, diagnosis report from yet another and consolidate both to produce a report and submit it to a third location). 

 

I am not sure if we are not handling such applications. But I feel we need to define a solid boundary for "application" type.

 

2) Checkpoint data management. Once the checkpoint data has been created, it needs to be managed. For example, there needs to functions for deleting checkpoint data that is no longer needed. Also, this API should provide a mechanism for querying the meta-data that is associated with checkpoint data.

 
I understand that Checkpoint has to be stored somewhere to be able to delete or query the meta-data. It is said tacitly in the above paragraph. But I feel we have to define the storage of checkpoint also as this is a big question in its own. This will have implications on Authentication, Authorization, Accounting, and also maintain the Data Integrity.
 
3) This is just a question I have. What happens when we have an application that takes checkpoint on its own and also Grid-CPR is checkpointing it. This may result in redundancy of meta data and wastage of CPU cycles. This sort of ties to Data Integrity in my previous comment. But we have to be able to recognize such applications and make a decision to checkpoint or not to checkpoint.
 
Regards,
Guru

		
---------------------------------
 Yahoo! Music Unlimited - Access over 1 million songs. Try it free.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.ogf.org/pipermail/gridcpr-wg/attachments/20051017/caae4a03/attachment.html