Date: Sun, 19 Nov 2000 23:45:50 -0500 (EST) From: David Miller <dmiller@search.sparks.net> To: Mike Silbersack <silby@silby.com> Subject: Re: UDP limits in dns server? Message-ID: <Pine.BSF.4.21.0011192332420.25661-100000@search.sparks.net> In-Reply-To: <Pine.BSF.4.21.0011192234220.59748-100000@achilles.silby.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, 19 Nov 2000, Mike Silbersack wrote: > On Sun, 19 Nov 2000, David Miller wrote: > > I've increased net.inet.udp.recvspace to 192k. Is there anything else I > > can do to tune the system? I'm particularly perplexed that a K6-200 > > system I had was cpu bound running named and achieved ~200 > > resolves/sec; my spiffy new 1100 MHz K7 is struggling to double it. > > > > Any suggestions welcome! > > > > --- David > > You may wish to try dan bernstein's dnscache resolver, it's part of the > djbdns port. (Actually, the port may still be named dnscache, I'm not > certain.) > > No clue if it's actually better. Dan says it is of course, and it's not > at all based on the bind resolver, so it'll certainly perform > different. I'm sure people would be interested to hear the results either > way. As a matter of fact, I am using it right now. I tried bind8.x and it worked OK, but seemed to slow down as the cache grew. I was surprised that the CPU could be a limit here, but it certainly seemed to be. I think the best run I had (256K addresses) was just shy of 400/sec. I got half that on 1/5 the mchine, so I went looking elsewhere. I'm trying the dnscache component of the djbdns package and it does have some nice features. You can set limits on the number of requests outstanding, how much memory for it to take, etc, and it works nicely. I haven't tried any other components of the system yet so I can't speak for those. The same run on dnscache seemed to go a bit over half the speed at 227 lookups/second. I've had the suggestion to up maxusers to 512. I'm a little leery of this, having heard of problems > 128 on this list for seemingly years. Is there a better way to size the network buffers and anything else which needs to be adjusted? Thanks Mike, Alfred, and all. --- David 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?Pine.BSF.4.21.0011192332420.25661-100000>