>> I really hope that this was some sort of glitch in your test rig. >> Could you please re-run this experiment a dozen times or so and post >> the results? > > I'll run it continuously overnight and post the results. But I'll scale > down from 40 storage nodes to 12; I think 10 nodes per machine was > excessive.
Here are the results. The variance is large but the general conclusion from earlier appears correct -- on my low-latency benchmarking grid, 1.7.1 is better for downloading large files and 1.8.0c3 is better for downloading small files. This result seems fairly consistent, but it's also not large relative to the run-to-run variance. 1.7.1: upload large 61.05 37.72 44.08 90.67 41.85 42.72 44.99 43.06 38.96 upload small 812.27 849.09 868.49 1398.67 814.20 803.51 826.16 808.24 775.21 download large 169.10 197.72 188.20 202.36 198.46 192.38 155.87 189.94 197.13 download small 255.39 250.01 253.62 248.31 246.43 250.32 253.09 247.13 245.02 1.8.0c3: upload large 41.27 41.29 42.50 42.12 53.32 40.65 49.33 38.67 40.80 upload small 822.21 811.20 853.65 868.70 843.07 854.33 808.97 774.10 790.48 download large 215.19 213.50 221.18 197.91 169.79 210.96 205.16 215.57 215.45 download small 218.77 222.93 332.70 222.75 223.13 225.54 221.65 212.61 212.50 -- Kyle Markley _______________________________________________ tahoe-dev mailing list [email protected] http://tahoe-lafs.org/cgi-bin/mailman/listinfo/tahoe-dev
