Date: Fri, 21 May 2004 11:15:33 -0400 (EDT) From: Andrew Gallatin <gallatin@cs.duke.edu> To: "Steven Hartland" <killing@multiplay.co.uk> Cc: freebsd-amd64@freebsd.org Subject: Re: Why is MySQL nearly twice as fast on Linux/AMD64 Vs.FreeBSD/AMD64? Message-ID: <16558.7445.195712.176174@grasshopper.cs.duke.edu> In-Reply-To: <00de01c43eeb$f2753220$b3db87d4@multiplay.co.uk> References: <200405202235.46755.Gregor.Bittel@GMX.de> <5.2.0.9.2.20040520122752.015eec60@mail.ojoink.com> <Pine.LNX.4.53.0405201129570.22287@sql01.internal.mikehost.net> <5.2.0.9.2.20040520141038.0432d130@mail.ojoink.com> <007701c43ec5$ff0a6fd0$b3db87d4@multiplay.co.uk> <00de01c43eeb$f2753220$b3db87d4@multiplay.co.uk>
next in thread | previous in thread | raw e-mail | index | archive | help
Steven Hartland writes: > Ok its IO (netcard drivers). > TBH after looking at this benchmarker I'd say its pretty useless > its testing packet throughput / rpc time not the database. > > Typical db load is not all insert statements its not even mostly > insert statements its selects. Saying that there's some startling results > below. > Exec summary: > FreeBSD 5.1.2 (i386) > local: 22834.872368421 records per second > fxp 100Mb: 3854.06863517 records per second > bge 100Mb: 2501.66999862 records per second > bge 1Gb: 3281.65229885 records per second Nice work. Does this translate pretty much directly into netperf request/response results? (netperf -tTCP_RR -Hremote) Assuming it does, have you tried "tuning" the bge interrupt coalescing params? Without adding an interface, I'd just suggest doing something like changing sc->bge_rx_max_coal_bds from 64 to 1 around like 2350 of dev/bge/if_bge.c, rebuilding the driver, and seeing what changes. I'm betting that will turn off recv interrupt coalescing and reduce your latency some. Drew
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?16558.7445.195712.176174>