Tom Lane wrote: > Perhaps it's time to remind people that what we want to measure > is the performance seen by a C program issuing write() and read() >commands, transferring 8K at a time, on a regular Unix filesystem
Yes...and at the risk of being accused of marketing ;-) , that is exactly what the 3 programs in my archive do (see previous post for url) : - one called 'write' creates a suitably sized data file (8k at a time - configurable), using the write() call - another called 'read' does sequential reads (8k at a time - configurable), using the read() call - finally one called 'seek' does random reads (8k chunks - configurable), using the lseek() and read() calls I tried to use code as similar as possible to how Postgres does its io....so the results *should* be meaningful ! Large file support in enabled too (as you need to use a file several times bigger than your RAM - and everyone seems to have >1G of it these days...) I think the code is reasonably readable too.... Its been *tested* on Linux, Freebsd, Solaris, MacosX. The only downer is that they don't automatically compute random_page_cost for you..(I was more interested in the raw sequential read, write and random read rates at the time). However it would be a fairly simple modification to combine the all 3 programs into one executable that outputs random_page_cost... regards Mark ---------------------------(end of broadcast)--------------------------- TIP 6: Have you searched our list archives? http://archives.postgresql.org
