Re: [Mimedefang] Socket map performance
David F. Skoll wrote: Have you used it for standard maps? (e.g. access or virtusertable) Yes. Same YES => Have you tried to reduce (horrible) number of lookups issued by "any map" sendmail.cf design? Jupp, with a little trickery it can be done one lookup for say www.noone.com is possible mostly. No. But a socketmap lookup passed through MIMEDefang down into the Perl code is something like 10x to 1000x slower than a Berkeley DB lookup. Ooo .. idd note to self .. find a way around this nasty beast, but .. as I only run a moderately small server :) cost gain benifits are well u know :) (Side note: I'd love to see Sendmail support Dan Bernstein's "CDB" databases; our tests indicate they are much faster than Berkeley DB. Maybe that's a project for the future...) Dit not look into that .. but will thnx -- Michiel Brandenburg ___ NOTE: If there is a disclaimer or other legal boilerplate in the above message, it is NULL AND VOID. You may ignore it. Visit http://www.mimedefang.org and http://www.roaringpenguin.com MIMEDefang mailing list MIMEDefang@lists.roaringpenguin.com http://lists.roaringpenguin.com/mailman/listinfo/mimedefang
Re: [Mimedefang] Socket map performance
Andrzej Adam Filip wrote: > Have you used it for standard maps? (e.g. access or virtusertable) Yes. > YES => Have you tried to reduce (horrible) number of lookups issued by > "any map" sendmail.cf design? No. But a socketmap lookup passed through MIMEDefang down into the Perl code is something like 10x to 1000x slower than a Berkeley DB lookup. (Side note: I'd love to see Sendmail support Dan Bernstein's "CDB" databases; our tests indicate they are much faster than Berkeley DB. Maybe that's a project for the future...) Regards, David. ___ NOTE: If there is a disclaimer or other legal boilerplate in the above message, it is NULL AND VOID. You may ignore it. Visit http://www.mimedefang.org and http://www.roaringpenguin.com MIMEDefang mailing list MIMEDefang@lists.roaringpenguin.com http://lists.roaringpenguin.com/mailman/listinfo/mimedefang
Re: [Mimedefang] Socket map performance
"David F. Skoll" wrote: > Michiel Brandenburg wrote: > >> What might be a nice tradeoff is using the socket map feature of >> sendmail to hook sendmail into mimedefang that way. Dunno how that >> would impact performance but it might be a nice tradeoff. > > Performance impact is severe. We used the sockemap -> MIMEDefang > path in an older version of our commercial software, but we had > to remove it and use a different technique because of the horrible > performance. Have you used it for standard maps? (e.g. access or virtusertable) YES => Have you tried to reduce (horrible) number of lookups issued by "any map" sendmail.cf design? >> As I recall the socket map protocol is pretty lightweight but might >> still be too slow, in my case it works fine even with about 1/2 mails >> per sec. > > We tend to concentrate our optimizations on medium-sized > installations, by which I mean about 25 msgs/second (= about 2 > million/day) or higher. The low-end ones aren't worth worrying about > just because the load is easily managed. -- [pl>en: Andrew] Andrzej Adam Filip : a...@onet.eu "My sense of purpose is gone! I have no idea who I AM!" "Oh, my God... You've.. You've turned him into a DEMOCRAT!" -- Doonesbury ___ NOTE: If there is a disclaimer or other legal boilerplate in the above message, it is NULL AND VOID. You may ignore it. Visit http://www.mimedefang.org and http://www.roaringpenguin.com MIMEDefang mailing list MIMEDefang@lists.roaringpenguin.com http://lists.roaringpenguin.com/mailman/listinfo/mimedefang