I ran sysbench on my old server, which is a Athlon64 at 1.8Ghz, with 1Gb of RAM and an older 100G disk, running at ATA33 because of the cable I have hooked up to it. So, pretty crappy. I ran sysbench with the random write and random read tests with 1, 2, and 3 threads each.
sysbench --test=fileio --file-test-mode=rndwr run sysbench --test=fileio --file-test-mode=rndrd run sysbench --test=fileio --num-threads=2 --file-test-mode=rndwr run sysbench --test=fileio --num-threads=2 --file-test-mode=rndrd run sysbench --test=fileio --num-threads=3 --file-test-mode=rndwr run sysbench --test=fileio --num-threads=3 --file-test-mode=rndrd run Here's a link to the Google Docs spreadsheet with the numbers: https://spreadsheets.google.com/ccc?key=0ArcmL0I6rCOEdDVZbXpCMDJBWkEzcTlVa0htNlJVYXc&hl=en And here's the graph as a separate image: http://leaf.dragonflybsd.org/~justin/sysbench_avg.png There's a huge improvement between 2.6 and 2.8, and some from 2.8 to 2.10. My guess would be that if I tried this on modern hardware - especially an AHCI-based system - the numbers would look even better. I tried using rebench but had a wierd error about a missing NumPy header file when compiling; I didn't get farther. (suggestions welcomed.)
