FWIW, I did some further experiments. Disabling mem-pool entirely (in favor of plain malloc/free) brought run time down to 3:35, vs. 2:57 for the exact same thing without multiplexing. Somehow we're still not managing contention very well at this kind of thread count, but the clues and opportunities are becoming less obvious. _______________________________________________ Gluster-devel mailing list Gluster-devel@gluster.org http://www.gluster.org/mailman/listinfo/gluster-devel
- [Gluster-devel] Multiplexing - good news, bad news... Jeff Darcy
- Re: [Gluster-devel] Multiplexing - good news,... Avra Sengupta
- Re: [Gluster-devel] Multiplexing - good n... Jeff Darcy
- Re: [Gluster-devel] Multiplexing - go... Avra Sengupta
- Re: [Gluster-devel] Multiplexing - good news,... Jeff Darcy
- Re: [Gluster-devel] Multiplexing - good n... Vijay Bellur
- Re: [Gluster-devel] Multiplexing - go... Jeff Darcy
- Re: [Gluster-devel] Multiplexing - good news,... Raghavendra Gowdappa
- Re: [Gluster-devel] Multiplexing - good news,... Xavier Hernandez
- Re: [Gluster-devel] Multiplexing - good n... Jeff Darcy
- Re: [Gluster-devel] Multiplexing - good news,... Pranith Kumar Karampuri
- Re: [Gluster-devel] Multiplexing - good n... Jeff Darcy