- 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

Sandia team changes have been deposited
by Ward, Lee 28 Feb '11

28 Feb '11

All, I have deposited the Sandia team's changes. Missing a couple of CVs but those are not likely to produce conflicts so I believe we can proceed. A couple of members on my team provided some approximate budget information. I did not modify the document to actually include that content. I dropped these budget files into the bib directory. Doesn't seem the right place but it's where I found budget info from others. I'm going to send a tar-ball of what I checked in back to my team for a double-check. I'm confident I got everything, though, and don't expect any oopsies. If so, I'll handle fixes amidst the nromal flurry of activity. In short, I believe we can remove our virtual lock on the document. --Lee

3 2

build broken?
by Phil Carns 18 Feb '11

18 Feb '11

Anyone know why the pdf won't build at the moment? -Phil

2 2

Fwd: Epidemic routing
by Rob Ross 16 Feb '11

16 Feb '11

FYI. Epidemic stuff from the other team. I haven't read it yet, but sounds relevant and like we have another fan of the approach. -- Rob Begin forwarded message: > From: Richard Brooks <rrb(a)ACM.ORG> > Date: February 16, 2011 6:55:25 PM CST > To: P2PFS-L(a)CLEMSON.EDU > Subject: Epidemic routing > Reply-To: PEER-TO-PEER FILE SYSTEM DISCUSSION <P2PFS-L(a)CLEMSON.EDU> > > Attached is an IEEE Trans SMC paper and Chapter 11 from > my Security Book (Disruptive Security Technologies with > Mobile Code and Peer-to-peer networks, by R. R. Brooks, > CRc Press, Boca Raton, 2005). > > I am most happy with the SMC paper, but included the Chapter > because it deals explicitly with the continuum: > > Flooding -- Gossip -- Random walk(s) > > There is a similar chapter in my distributed sensor > networks book. > > Basically, random graph theory in math mainly looks for the > giant component threshold in graphs as the number of nodes > becomes infinite. The physics counterpart finds the percolation > threshold in a regular grid as the size of the grid becomes > infinite. > > You can use similar principles for finite systems. The mobile > net SMC paper shows how to do some of this. For gossip > protocols and n random walkers you can calculate the probability > distributions. The percolation thresholds I find are pretty robust, > and we have done this to do a number of interesting things: > clone detection, key server assignment, etc. > > Got to go back to proposal writing. Would be glad to expound on > this approach until everyone else falls asleep. > -- > =================== > R. R. Brooks > > Associate Professor > Holcombe Department of Electrical and Computer Engineering > Clemson University > > 313-C Riggs Hall > PO Box 340915 > Clemson, SC 29634-0915 > USA > > Tel. 864-656-0920 > Fax. 864-656-5910 > email: rrb(a)acm.org > web: http://www.clemson.edu/~rrb >

1 0

Lee's "full pass"
by Ward, Lee 15 Feb '11

15 Feb '11

I did as I said I would. Changed very little. Some small sentence wording, due to awkward construction. Changed a little to use less, um, judgemental. I note that the sections at the end, like risk mitigation, have had no love yet. As I put in my comment, there, I don't think we want this in the proposal. --Lee

2 1

Can't tag the repository :(
by Ward, Lee 15 Feb '11

15 Feb '11

I cannot tag the proposal. It doesn't work the way CVS does and our repository does not appear to be set up to support the equivalent, copy, operation. Instead, I recorded the revision. I shipped revision number 196. --Lee

2 1

assignments
by Rob Ross 15 Feb '11

15 Feb '11

Getting close, thanks! -- Rob --- RobR: pass over 7.x, take page counts out, cut todos that i can. Phil: pass over 7.2 milestones, no more team language Sam: pass over 6.x Galen: pass over Exec. Summary Lee: full pass Phil: pass over 4.x Phil: improve titles of 4.1 and 4.2 to be more clear about reads and writes. Phil: heterogeneous into 4.3 title. Scott: pass over 4.x after Phil All: Get CVs in. Fix refs.

2 1

assignments
by Rob Ross 15 Feb '11

15 Feb '11

RobR: 5.0 (match with 3.0 and 4.0 tone), 5.1 (include geodesic figure here), 2.4 RobR: 4.0, title change: Placing data to improve performance Dries: 2.3 Sam: 3.1, active and passive models, may need some more information from Lee Lee: 3.3, 4.3, look, improve. Brad: 5.3 Scott: 3.2, trim down some, maybe "redundant storage for scientific workloads" Phil: 3.3 Galen, Phil: Read through, get a feel for level of detail, etc. --- Also, Friday concall at 2-3pm Eastern. Rob

4 35