Date: Thu, 31 Jan 2002 17:33:28 +0000 (GMT) From: Mike Silbersack <silby@silby.com> To: Storms of Perfection <gary@outloud.org> Cc: thierry@herbelot.com, <replicator@ngs.ru>, <hackers@FreeBSD.ORG> Subject: Re: Clock Granularity (kernel option HZ) Message-ID: <20020131172729.X38382-100000@patrocles.silby.com> In-Reply-To: <3197.208.141.46.249.1012516570.squirrel@test.outloud.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, 31 Jan 2002, Storms of Perfection wrote: > I'm going to benchmark different network senarious with different options > to see what I can get, and what works best. If someone wants to help me > out, I could maybe write up a article about it? I don't think you'll end up seeing the performance improvements you're looking for. The case where HZ=1000 is really useful is when using dummynet; the more accurate scheduling is necessary for it to handle high data rate pipes properly. The TCP stack, on the other hand, is perfectly happy with 10ms resolution. Retransmission timeouts are only actually used when loss occurs on the network, and 10ms is more than accurate enough for retransmission. (I believe that retransmit timeouts are rounded up to 1 second, but don't quote me on that.) The other timed events (keepalive timeouts, delayed ack timeouts, etc) are also in good shape with 10ms accuracy. So, it's highly unlikely that you'll be able to observe a perceptable difference in network performance except in really convoluted cases. Mike "Silby" Silbersack To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20020131172729.X38382-100000>