Return to BSD News archive
Path: euryale.cc.adfa.oz.au!newshost.carno.net.au!harbinger.cc.monash.edu.au!munnari.OZ.AU!spool.mu.edu!uwm.edu!cs.utexas.edu!howland.erols.net!news.mathworks.com!uunet!in1.uu.net!twwells!twwells!not-for-mail
From: bill@twwells.com (T. William Wells)
Newsgroups: comp.unix.bsd.freebsd.misc
Subject: Re: FreeBSD as news-server??
Date: 15 Oct 1996 17:19:32 -0400
Organization: None, Mt. Laurel, NJ
Lines: 51
Message-ID: <540v54$ap@twwells.com>
References: <537ddl$3cc@amd40.wecs.org> <53u1ic$61i@flash.noc.best.net> <53ucuj$8qh@twwells.com> <DzBp9x.FwD@interactive.net>
NNTP-Posting-Host: twwells.com
In article <DzBp9x.FwD@interactive.net>,
Chris Mauritz <ritz@interactive.net> wrote:
: My setup is like this:
: [nice setup :-)]
Just for reference, my own setup is this:
FreeBSD2.1.5, on a 100MHz Pentium, 64M RAM, Adaptec 2940
controller. Disks:
FUJITSU M2694ES-512 812A (1G) System stuff, swap, news library
CONNER CFP4207S 4.28GB 1420 article spool (everything else)
overview, in.coming, out.going, etc.,
CONNER CFP4207S 4.28GB 2B4B article spool (most alt, big 8, uk)
CONNER CFP4207S 4.28GB 1420 article spool (alt.binaries)
4 incoming newsfeeds (via a pair of T1s).
2 outgoing newsfeeds (not counting the tiny ones).
I tend to have 15-20 nnrpds at most times; ~40 at peak times.
For software, I started with INN1.4unoff4. These are the major
performance related modifications:
o Incoming newsfeeds are routed to a small nntp daemon
optimized to minimize latency. (One incoming stream, one
daemon. No streaming.)
o Duplicate data in the incoming spool is minimized by a
msgid daemon that keeps 4 hours of message ids around.
o Nnrpds are spawned from a connectd descendent.
o Innd reads files directly from in.coming. (So it's now a
fancy relaynews. :-)
o Article directories have subdirectories with no more than
1000 articles/directory.
o Increased sizes of various data structures. (E.g.,
initial list sizes, hash table buckets.)
I have a dynamic expiration policy: warez groups are kept .25
days, binaries .25 or 2 days, all others 2 or 25 days. The longer
periods apply if someone's been reading the group in the last two
weeks. I have over 1100 groups that are kept for 25 days.
With this setup, I get 3-3.5 articles/second, depending on the
nnrp load. (Bare unoff4 on an unloaded system only did around 2.)
I occasionally fall behind, perhaps by an hour, if something hits
the disks particularly badly.