March 2011 - X-store - mailman3.cels.anl.gov

comments on 5.2, discovery of data
by Phil Carns 23 Aug '11

23 Aug '11

I don't know if there will end up being room or not, but when I was reading this I was thinking that it would be neat to have a figure there to illustrate the idea of constructing multiple name spaces out of attribute/provenance information. Maybe something with three panels; the first would list some objects and their attributes just in a plain table. Something like this, following the examples discussed in the text (maybe labeling the categories of standard provenance vs. domain-specific attributes): object jobid owner ocean_mod_resolution -------------------------------------------- A 123 user1 5000 B 123 user1 5000 C 222 user1 N/A D 245 user2 4000 E 245 user2 4000 F 245 user2 4000 The other two panels would show a couple of name spaces created from those objects; one with a tree hierarchy that organizes the objects by owner and then job id, and the other with a tree that organizes by ocean model resolution (and leaves out object C, which was from a different problem domain). Might help illustrate how its possible to have multiple hierarchical name spaces overlaying the same data. I'd be happy to help draw something like that if needed (and if anyone else thinks it would be a good idea). BTW, I have no idea what range of values make sense for ocean model resolution :) My only comment on the text itself is that I think the last couple of sentences of the first paragraph (about the implementation challenge of how to maintain indexes) could be moved to the last paragraph instead. I think that point fits in better with the other implementation-y stuff at the end than it does with the introductory text. Also makes sure that the idea sounds interesting before hinting that it might be hard to build :) -Phil

2 5

some updates
by Rob Ross 04 Mar '11

04 Mar '11

hey, i've hacked on the text provided by our NNSA collaborators quite a bit. mainly it was present in the networking and security subsections in Core. It's back down to size now. I've also tried a new title out: Revolutionizing Storage for Exascale Science. I think it's more compelling, but I'm open to suggestions. I plan to make a pass through some of the other sections later today, but I thought I'd give you a heads up about the changes so far. It's beginning to look like a real proposal! There has also been some progress on a possible funding avenue, but I want to get a little more confidence before I make any claims :). I'll keep you posted. We should plan another call early next week, or maybe to meet in San Diego for those that will be there? Who is coming? Thanks, Rob

4 4

bibtex
by Phil Carns 02 Mar '11

02 Mar '11

I made a pass through the document today and filled in the undefined citations and references. The only one left is the \cite{EXASCALESSD} one in the intro. I wasn't sure what to reference for that one. -Phil

2 1