Date: Wed, 7 Jul 1999 23:13:08 -0500 From: Jonathan Lemon <jlemon@americantv.com> To: Mohit Aron <aron@cs.rice.edu> Cc: freebsd-net@freebsd.org Subject: Re: paper on improving webserver performance Message-ID: <19990707231308.40142@right.PCS> In-Reply-To: <199907080346.WAA27087@cs.rice.edu>; from Mohit Aron on Jul 07, 1999 at 10:46:19PM -0500 References: <199907080340.WAA29430@free.pcs> <199907080346.WAA27087@cs.rice.edu>
next in thread | previous in thread | raw e-mail | index | archive | help
On Jul 07, 1999 at 10:46:19PM -0500, Mohit Aron wrote: > > > > This seems to indicate that you still have the overhead where the timer > > fires, but no events are actually pending. > > True, but it avoids any pointer manipulations when timers are set and cancelled > by TCP (which happens much more). Only if a timer is set by TCP that needs to > fire earlier than the time that the timing wheel event handler needs to fire > would you need to move around the timing wheel event. Yes, I thought the same thing too. Originally, my implmentation simply moved all connections in a TIME_WAIT states onto a separate timing wheel, and left the rest of the code alone. Turns out that the overhead of inserting/deleting events from the timing wheel is trivial as compared to the cost of checking the hash buckets and walking the chains checking for events. > In any case, I'm very interested in your implementation since you appear to > have seen faster performance results with it. I'm just about done; wrapping up some testing and making sure I didn't break anything. -- Jonathan To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-net" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?19990707231308.40142>