People call me persistent too. As in "persistent nuisance". Re: Time Warm and optimistic schedulers: I'm not a big fan of them, at least the the classes of problems that I usually end up simulating. The systems are too tightly coupled for the scheduler to show any advantage. In fact, they end up doing endless rollbacks. I have, however implemented an efficient random pair-wise synchronization algorith for running on distributed hardware. I described it to someone else on the list earlier:
EpiSims uses a relatively new method of synchronization that does not require a centralized controller: each slave cpu in the run randomly synchronizes with other slave processes during the run. This is done in such a way that the entire machine stays in sync without a centralized controller; therefore, it scales well. This Slashdot article discusses this style of parallel synchronization. The Slashdot article includes a link to the NewScientist.com article where I originally read about the approach in 2003. http://slashdot.org/articles/03/02/01/1342203.shtml?tid=126 These references discuss the specific approach that I implemented in EpiSims. I discovered that one of the authors was Zoltan Toroczkai who also worked at LANL at the time. It is a difficult algorithm to implement, but as the articles indicate it is worth the effort because of the improved performance and scaling behavior of the distributed application as compared to centralized controller synchronization approaches. http://www.trnmag.com/Stories/2003/021203/Scheme_smooths_parallel_processing_021203.html http://www.rpi.edu/~korniss/Research/gk_research.html<http://www.rpi.edu/%7Ekorniss/Research/gk_research.html> --Doug On Tue, Sep 1, 2009 at 7:05 AM, Russ Abbott <[email protected]> wrote: > It was interesting to watch myself fish for the names WarpIV and Jeff > Steinman when I replied to Doug's message. > > I knew that Doug's description reminded me of a system I had known about. > But I couldn't retrieve either WarpIV or Jeff Steinman from my own memory > even though I knew it was that system that I wanted to mention to Doug. How > to find those names. I knew that Jeff's company was in San Diego. I also > knew that he had previously worked at JPL. So I did a search for JPL, > Simulation, and San Diego. Among the items returned, the term "time warp" > showed up. I recognized that as part of Jeff's approach. So I added that to > the search. That search returned an item with the name WarpIV. I recognized > that and tracked down Jeff and his company. > > What was interesting was that I knew what I wanted, I didn't know its name, > but I was able to recognize relevant terms when Google thrust them in front > of me. That and some persistence was what it took. > > -- Russ Abbott > _____________________________________________ > Professor, Computer Science > California State University, Los Angeles > Cell phone: 310-621-3805 > o Check out my blog at http://russabbott.blogspot.com/ > > > ============================================================ > FRIAM Applied Complexity Group listserv > Meets Fridays 9a-11:30 at cafe at St. John's College > lectures, archives, unsubscribe, maps at http://www.friam.org > --
============================================================ FRIAM Applied Complexity Group listserv Meets Fridays 9a-11:30 at cafe at St. John's College lectures, archives, unsubscribe, maps at http://www.friam.org
