Pegasus meeting notes
From CEDPS
Contents |
A collection of meeting notes from discussions with the Pegasus team.
Teleconference on 7/14/08
Attending: Dan, Keith, Binit, Brian from CEDPS; Ewa and Karan from ISI
Upcoming posters/papers
SC poster, July 31:
- use old SCEC data sets
- system integration
- scalability of workflow system and analysis system
eScience, Aug. 10:
- SCEC-centric, use new results
- ISI will lead
- topics: workflow progress tool, error analysis, perf. analysis
WASL Usenix, Sep. 2:
- infrastructure focused
- use Pegasus/SCEC as example
- LBL will lead
Farther out, HPDC and Scientific and Statistical Database Management Conference
Action items
* Ewa will send out SciDAC poster paper outline by tomorrow -- then, we'll do task assignment * Outlines for eScience by end of week (also Ewa) * Gaurang will re-send some netlogger db query issues
Teleconference on 6/5/08
Attending (from memory): Dan, Keith, Binit from CEDPS; Ewa, Gaurang, Paul from ISI.
CEDPS team introduced Binit.
Gaurang talked about load time, query time. Discussed possibly building temporary tables for querying.
w.r.t. next steps, Gaurang will send some logs w/Dagman data that is easily parsed. CEDPS team will try parsing & playing with that.
Talked about Binit doing web page. Pegasus team would prefer command-line scripts. So Binit's task was modified to first write a command-line program to do the "basic queries" Ewa sent out a while back. Initially, just getting text results. He can still build the web page on top of that, later.
Ewa mentioned that comparing Dagman logs and Kickstart records may provide interesting performance insights since Dagman runs "clusters" of jobs and the comparative performance of those clusters will indicate how good their clustering algorithm is.
There was a question whether this ends up in the netlogger repo under pegasus/ or the pegasus repo under netlogger/ (to paraphrase). Agreed put it in the pegasus repo; they're ok with that; Gaurang will give us svn access.
No paper targets were picked. But agreed that an SC poster, which has a July 31 abstract deadline, might be a good idea.
