From strenholme.usenet at gmail.com Mon Oct 1 13:47:45 2007 From: strenholme.usenet at gmail.com (Sam Trenholme) Date: Mon, 1 Oct 2007 17:47:45 +0000 Subject: MaraDNS Large Sites? In-Reply-To: References: Message-ID: <7bd685720710011047j6ce75d38m83fdba778bfbe14d@mail.gmail.com> Are you going to use MaraDNS as an authoritative or recursive DNS server. > I'm considering deploying MaraDNS for a site that would see around 300 DNS > requests per second. MaraDNS, on a modern computer should easily be able to process this many authoritative records: http://www.maradns.org/speed.comparison.html For recursion, this may be pushing MaraDNS. I'm going to rewrite the recursive core in a little bit to no longer use threads; until then, make sure you do workload tests when having a highly loaded recursive DNS server. - Sam From duane at e164.org Mon Oct 1 19:12:20 2007 From: duane at e164.org (Duane) Date: Tue, 02 Oct 2007 09:12:20 +1000 Subject: MaraDNS Large Sites? In-Reply-To: <7bd685720710011047j6ce75d38m83fdba778bfbe14d@mail.gmail.com> References: <7bd685720710011047j6ce75d38m83fdba778bfbe14d@mail.gmail.com> Message-ID: <47017ED4.7030807@e164.org> Sam Trenholme wrote: > Are you going to use MaraDNS as an authoritative or recursive DNS server. > >> I'm considering deploying MaraDNS for a site that would see around 300 DNS >> requests per second. > > MaraDNS, on a modern computer should easily be able to process this > many authoritative records: > > http://www.maradns.org/speed.comparison.html Any plans to do a comparison testing on the current versions of MaraDNS (auth and recursion), BIND (auth and recursion), djbdns (auth and recursion), NSD (auth only) and PowerDNS (auth or recursion), and any others that have released at least a security fix in the last 6-12 months. I don't want this to sound like a rant, or that I'm out to get Sam in anyway, but things change all the time. However I can't help but notice the versions listed on the link given are quite old, even djbdns which hasn't been updated in years released you have compared against a previous minor version then what's currently distributed. The closest thing I can find to the above criteria was on a Hungarian site (but I don't speak/read Hungarian and I can't find a decent online translator): http://www.gluon.hu/dns_benchmark They used MaraDNS v1.2.07.5 and djbdns 1.0.5, while not 100% current is a little closer to current then 0.5.09. From jp at mens.de Tue Oct 2 02:58:09 2007 From: jp at mens.de (Jan-Piet Mens) Date: Tue, 2 Oct 2007 08:58:09 +0200 Subject: MaraDNS Large Sites? In-Reply-To: <47017ED4.7030807@e164.org> References: <7bd685720710011047j6ce75d38m83fdba778bfbe14d@mail.gmail.com> <47017ED4.7030807@e164.org> Message-ID: <20071002065809.GA28231@home.mens.de> On Tue Oct 02 2007 at 01:12:20 CEST, Duane wrote: > Any plans to do a comparison testing on the current versions of MaraDNS > (auth and recursion), BIND (auth and recursion), djbdns (auth and > recursion), NSD (auth only) and PowerDNS (auth or recursion), and any > others that have released at least a security fix in the last 6-12 months. I plan on doing a performance comparison soon. I had planned on creating 100.000 domains, each with about 5-10 RR and having queryperf with a fixed set of queries run against all of the above. I'd like to add a chapter on performance for my book, comparing the individual servers. -JP From blist at mikkl.com Tue Oct 2 05:51:17 2007 From: blist at mikkl.com (blist) Date: Tue, 2 Oct 2007 05:51:17 -0400 Subject: MaraDNS Large Sites? In-Reply-To: <47017ED4.7030807@e164.org> References: <7bd685720710011047j6ce75d38m83fdba778bfbe14d@mail.gmail.com> <47017ED4.7030807@e164.org> Message-ID: <200710020551.17345.blist@mikkl.com> On Monday 01 October 2007, Duane wrote: > |The second column seems to be memory usage, MaraDNS seems to use the > |most, any way to improve this to a comparable level of BIND9? Although > |with ram/hardware prices going down all the time this might not be that > |much of an issue, unless you are in a virtual hosting situation with 64M > |of ram or even 128 might start to matter depending how many zones etc > |your are planning to host. I looked at those results and then compared it to my local configuration of maradns as a recursive server supporting a single machine. According to KDE System Guard, mara is only taking 34 Meg of virtual memory and is noticeably less heavy on my systems than bind. Yes, I realize this may be comparing apples and oranges, but thought I would share the data none-the-less. -- mikkl PCLinuxOS 2007 Registered Linux User #430720 From strenholme.usenet at gmail.com Tue Oct 2 10:55:29 2007 From: strenholme.usenet at gmail.com (Sam Trenholme) Date: Tue, 2 Oct 2007 09:55:29 -0500 Subject: MaraDNS Large Sites? In-Reply-To: <47017ED4.7030807@e164.org> References: <7bd685720710011047j6ce75d38m83fdba778bfbe14d@mail.gmail.com> <47017ED4.7030807@e164.org> Message-ID: <7bd685720710020755y7d95d5abx3bf521e39cf6a96b@mail.gmail.com> > Any plans to do a comparison testing on the current versions of MaraDNS > (auth and recursion), BIND (auth and recursion), djbdns (auth and > recursion), NSD (auth only) and PowerDNS (auth or recursion), and any > others that have released at least a security fix in the last 6-12 months. We'll see. My plate is pretty full right now; I don't think too much has changed since that test was done seven years ago. > I don't want this to sound like a rant, or that I'm out to get Sam in > anyway, but things change all the time. Oh, I agree 100%. Read me djbdns rant: http://www.maradns.org/advocacy.html#djbdns > The second column seems to be memory usage, MaraDNS seems to use the > most, any way to improve this to a comparable level of BIND9? Yes. Use default zonefiles. :) http://www.maradns.org/tutorial/default_zonefile.html Basically, right now my emphasis is supporting MaraDNS, testing the upcoming 1.3.07 stable release, and rewriting the recursive half of MaraDNS. Then we can talk about improving the authoritative half. Thank you for your MaraDNS feedback. - Sam