Date: Mon, 11 Mar 1996 16:50:49 -0500 (EST) From: Brian Tao <taob@io.org> To: Joe Greco <jgreco@solaria.sol.net> Cc: mtaylor@cybernet.com, hackers@FreeBSD.ORG Subject: Re: Proper FreeBSD news machine Message-ID: <Pine.BSF.3.91.960311162746.29052Q-100000@cabal.io.org> In-Reply-To: <199603062336.RAA04112@solaria.sol.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, 6 Mar 1996, Joe Greco wrote:
>
> No. In particular look at the size of history.pag as a clue to how much RAM
> you need. My many-years-tried-and-true INN RAM formula is:
>
> Memory: 16MB + sizeof(history.pag) * 2MB + numclients * 1MB + numfeeds * 1MB.
That would be a very "safe" assumption to make if you plan on
running the sharedactive patch with INN's in.nnrpd. Without the
patch, I find that in.nnrpd chews up sizeof(active+newsgroups), which
can exceed 2MB if you have 20,000+ groups. With the patch, you can
easily get away with "0.5MB * numclients".
> > C) would separate SCSI busses help? (I plan to put a second 4.3Gb HD
> > in for the rest of the news spool)
>
> Go PCI SCSI if you can. Also, the more disks, the merrier (I have 14 but
> then I'm a performance freak).
Yes, definitely. I've got 9 drives here and they all blink
merrily away every second of the day, every day. Looks really nifty
in a darkened machine room. :)
> > D) whose SCSI card has the 'best' performance?
>
> I've had good luck with the AHA3940 and NCR-810 based cards. The AHA2940
> should work well too.
I've got three NCR's here. No problems.
> Pay close attention to the memory advice. I see so many people try to get
> by without enough memory. It doesn't pay. I run Exec-PC's news operation
> and they try to squeeze 150-200 nnrp clients onto a box with 128MB RAM.
> They complain to me that it "takes forever to connect". I wonder why. ;-)
Bullpuckey. :) trn over a local Ethernet connection to the news
server here takes less than 5 seconds to display the first newsgroup
prompt. This is with a 28MB in-core innd process a dozen crossfeeds
and 230 nnrpd's.
load averages: 0.70, 0.78, 0.82 16:43:33
271 processes: 2 running, 274 sleeping
Cpu states: 12.6% user, 0.8% nice, 16.9% system, 9.6% interrupt, 60.2% idle
Memory: 92M Active, 868K Inact, 16M Wired, 13M Cache, 176K Free
Swap: 262M Total, 252M Free, 4% Inuse
PID USERNAME PRI NICE SIZE RES STATE TIME WCPU CPU COMMAND
12957 news -6 4 28M 28M sleep 10:22 19.00% 19.00% innd
16183 root 40 0 648K 992K run 0:00 4.92% 0.69% top
16179 news 2 4 276K 1224K sleep 0:00 1.97% 0.65% in.nnrpd
16181 news 2 4 292K 760K sleep 0:00 2.21% 0.57% in.nnrpd
8614 news 2 0 232K 540K run 1:01 0.50% 0.50% innxmit
13973 news 2 0 304K 1272K sleep 0:09 0.27% 0.27% in.nnrpd
[...]
BTW, news.io.org is open to all sites for reading (posting is
restricted to io.org and our customer sites). I have a cron job that
checks on the number of connected readers, and I don't think I've seen
it drop below 150 in the past week.
--
Brian Tao (BT300, taob@io.org)
Systems Administrator, Internex Online Inc.
"Though this be madness, yet there is method in't"
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.3.91.960311162746.29052Q-100000>