Date: Wed, 17 May 2006 13:37:23 +0200 From: Dimitry Andric <dimitry@andric.com> To: Dmitry Pryanishnikov <dmitry@atlantis.dp.ua> Cc: Brett Glass <brett@lariat.org>, stable@freebsd.org, Colin Percival <cperciva@freebsd.org> Subject: Re: 4.11 snapshots? Message-ID: <446B0AF3.1070302@andric.com> In-Reply-To: <20060517130629.T64952@atlantis.atlantis.dp.ua> References: <200605160135.TAA04838@lariat.net> <57d710000605151942p2461338au561269fc5937aee7@mail.gmail.com> <7.0.1.0.2.20060515225038.08d72690@lariat.org> <446981CD.5000309@gmail.com> <4469C668.2060807@rerowe.com> <7.0.1.0.2.20060516104907.08788ad8@lariat.org> <446A0608.10608@freebsd.org> <20060517130629.T64952@atlantis.atlantis.dp.ua>
next in thread | previous in thread | raw e-mail | index | archive | help
Dmitry Pryanishnikov wrote: > Well, have you seen my simple performance benchmarking RELENG_4 vs 6? > IMHO it mimics quote common usage pattern: it just downloads a large file > with 10Mbps rate and stores it on UFS filesystem. On the same hardware > (i386 uniprocessor Celeron-333 system with 128Mb RAM and fast SAMSUNG > SP0802N > HDD using UDMA33) under the same conditions, using more optimal config > (INVARIANTS removed) RELENG_6 (and 5) _still_ uses >= 50% of CPU time > for (Intr+Sys), while RELENG_4 doesn't use more than 28% for them. Just as a test for RELENG_6, could you try setting kern.hz="100" in your loader.conf, and repeating your tests? I'm just guessing, but maybe the higher interrupt rate is a bit too much for an old Celeron... :) Cheers, Dimitry
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?446B0AF3.1070302>