Ahead of the community meeting (barely), here's where things are. I've been in meetings all week, so progress has been a bit slower than usual. When I've been able to do direct work at all, it has mostly been on testing and mostly for the prerequisite io-threads patch (http://review.gluster.org/#/c/15643/). Last night I was able to finish running perf-test.sh as Pranith had asked, ostensibly to see if there were any performance regressions in that over-long, single-threaded, totally unbalanced mix of operations. There weren't. In fact, there seemed to be a very slight improvement; I'll send out details later. In a more rigorous test, and especially with multiplexing enabled, the difference is like night and day in favor of the new version. Thanks to Pranith and Xavi for help with the patch.
Next week I expect (hope) to get back to multiplexing itself, hopefully with another round of performance/scalability numbers before the next meeting. _______________________________________________ Gluster-devel mailing list [email protected] http://www.gluster.org/mailman/listinfo/gluster-devel
