From owner-freebsd-current Tue Oct 29 10:28:32 1996 Return-Path: owner-current Received: (from root@localhost) by freefall.freebsd.org (8.7.5/8.7.3) id KAA15042 for current-outgoing; Tue, 29 Oct 1996 10:28:32 -0800 (PST) Received: from brasil.moneng.mei.com (brasil.moneng.mei.com [151.186.109.160]) by freefall.freebsd.org (8.7.5/8.7.3) with ESMTP id KAA15037; Tue, 29 Oct 1996 10:28:29 -0800 (PST) Received: (from jgreco@localhost) by brasil.moneng.mei.com (8.7.Beta.1/8.7.Beta.1) id MAA24642; Tue, 29 Oct 1996 12:27:22 -0600 From: Joe Greco Message-Id: <199610291827.MAA24642@brasil.moneng.mei.com> Subject: Re: Request to add this to FAQ re: swap space To: rgrimes@gndrsh.aac.dev.com (Rodney W. Grimes) Date: Tue, 29 Oct 1996 12:27:22 -0600 (CST) Cc: dyson@freebsd.org, dtc@scrooge.ee.swin.oz.au, current@freebsd.org In-Reply-To: <199610290356.TAA00668@GndRsh.aac.dev.com> from "Rodney W. Grimes" at Oct 28, 96 07:56:33 pm X-Mailer: ELM [version 2.4 PL24] Content-Type: text Sender: owner-current@freebsd.org X-Loop: FreeBSD.org Precedence: bulk > > I agree that more swap is better, but at least we need to KILL the notion > > that 2 X RAM is enough (IMO, 2 X RAM is never enough)!!! > > IMOH, and IMPO, 2 x ram is typically all the swap you should need, > if you go beyond this need for swap your machine is under memoried > and page faulting heavily. There are exceptions to this (things that > manipulate massive data arrays in malloc regions (yes, I have a client > with close to 4G of swap, but he also has 256MB of memory, which is > maxed for his motherboard), wuarchive.cdrom.com running 1000's of processes, > etc... I am still waiting to hear how to get 256MB RAM on a motherboard reliably. :-) Incidentally, I meant to see if I could get you a bit irritated by pointing out that the ASUS P/E-P55T2P4D motherboard specs state... System Memory Eight 72-pin SIMM Sockets Support 8MB to 512MB Use 4/8/16/32/64MB 72-pin DRAM Module with 70ns Fast Page Mode or EDO DRAM (Require Parity Memory to Support ECC Function) (60 ns DRAM required when using 66MHz external clock) I can't get anything more than 192MB to work reliably on this board (6 x 32MB SIMM modules) - unless we go with the mega mondo $$$$ 64MB SIMM's.. and even those I have doubts about. We put the 4 that we have on a P/I-P55T2P4 board and it refused to work until we relaxed the RAM timing to 70ns. Gives me a sick feeling. I would really like to get 512MB RAM on one of these boards. :-) > My personal machines never have more than 2xRAM as swap, if I run out > of this much swap I add more memory and resize swap to 2X: > > GndRsh:sales {102} swapinfo > Device 1K-blocks Used Avail Capacity Type > /dev/sd0b 32768 13376 19328 41% Interleaved > GndRsh:sales {103} > SkyRsh# swapinfo > Device 1K-blocks Used Avail Capacity Type > /dev/sd0b 32768 2380 30324 7% Interleaved > SkyRsh# > > Both 16MB machines, both under utilizizing swap under typical loads, > both machines started out 3 or 4 years ago as 4MB machines, both > grown over time as the code has become larger (:-(), and the work > load increased (:-)). > > Ohhh... and for real server class machines if I swap, I add memory > and run swap == ram size. (From my Netware days, realize Netware > requires you to have physical ram for all NLM's, otherwise you get > the nasty out of memory error, module not loaded). I respectfully disagree... a real server class machine CERTAINLY needs to have sufficient RAM to run its appointed tasks... but on the other hand, you probably do not want FreeBSD going around killing your processes because something unexpected happened that you did not expect. This box has sufficient RAM: load averages: 1.74, 1.47, 1.36 12:16:28 65 processes: 4 running, 61 sleeping Cpu states: 63.0% user, 13.2% nice, 19.8% system, 3.1% interrupt, 0.8% idle Mem: 176M Active, 1228K Inact, 21M Wired, 53M Cache, 7021K Buf, 300K Free Swap: 614M Total, 128K Used, 614M Free PID USERNAME PRI NICE SIZE RES STATE TIME WCPU CPU COMMAND 16420 news 84 0 43M 43M RUN 0:16 42.59% 37.88% expire 13487 news 79 4 101M 99M RUN 12:10 30.82% 30.82% innd 13491 news 61 0 5788K 6004K RUN 7:24 14.34% 14.34% innfeed 16326 news 2 0 244K 620K select 0:01 1.39% 1.34% innxmit 13494 news 2 0 12M 11M select 0:07 0.65% 0.65% innfeed But I have seen times where INN blows up to 180MB and several innfeed processes blow up to 40MB or so, and all of a sudden we are swapping. Do I want it to continue running? YES.... Do I want to rely on swap as a normal part of operations? NO.... but I like being able to fall back if I need to! ... Joe ------------------------------------------------------------------------------- Joe Greco - Systems Administrator jgreco@ns.sol.net Solaria Public Access UNIX - Milwaukee, WI 414/546-7968