From owner-freebsd-isp Wed Aug 16 7:39:11 2000 Delivered-To: freebsd-isp@freebsd.org Received: from unix.megared.net.mx (megamail.megared.com.mx [200.52.207.52]) by hub.freebsd.org (Postfix) with ESMTP id EE44337BFF2 for ; Wed, 16 Aug 2000 07:39:06 -0700 (PDT) (envelope-from ales@megared.net.mx) Received: from ales (ales.corp.megared.net.mx [200.52.193.2]) by unix.megared.net.mx (8.9.3/8.9.3) with SMTP id JAA39805; Wed, 16 Aug 2000 09:39:05 -0500 (CDT) (envelope-from ales@megared.net.mx) Message-ID: <00e901c0078f$a7ccaba0$02c134c8@megared.net.mx> From: "Alejandro Ramirez" To: "Dave Wilson" Cc: References: <004501c006d7$d4c06dd0$112821c4@sai.co.za> <045501c00701$d94ae5e0$02c134c8@megared.net.mx> <00af01c0075d$1a450de0$112821c4@sai.co.za> Subject: Re: [SQU] failrly big cache for lots of clients, extra tips ? Date: Wed, 16 Aug 2000 09:38:34 -0500 X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 5.50.4133.2400 X-MimeOLE: Produced By Microsoft MimeOLE V5.50.4133.2400 Sender: owner-freebsd-isp@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.org Hi Dave, Squid doesnt use excessive bandwith, your users do, squid only intercepts the calls and makes the retrieve of files for your users, if you configure more disk for caching, your users will use less bandwith each time, but you will certanly increase the amount of RAM in your system. Greetings... Ales ----- Original Message ----- From: "Dave Wilson" To: "Alejandro Ramirez" Cc: Sent: Wednesday, August 16, 2000 3:36 AM Subject: Re: [SQU] failrly big cache for lots of clients, extra tips ? > Hi Alejandro, > > Thanks for those tips I will give it a try. > Also any ideas on how to get squid to not use excessive bandwidth ? > > Regards > Dave Wilson > The S.A. Internet > (033) 3456777 > 0825496159 > http://www.sai.co.za > "Who is General Failure and why is he reading my hard drive ?" > > ----- Original Message ----- > From: "Alejandro Ramirez" > To: "Dave Wilson" > Sent: Tuesday, August 15, 2000 11:43 PM > Subject: Re: [SQU] failrly big cache for lots of clients, extra tips ? > > > > > Any ideas ? > > > > Lots of them :o) > > > > Optimization Tips for Increase the Server Performance Amazingly > > > > Try enabling this options in your Makefile (squid port), to improve > > performance > > > > # - Optimize time updates to one per second rather than calling > > gettimeofday() > > CONFIGURE_ARGS+= --enable-time-hack > > # - Compile out code that does optional Ident (RFC931) lookups > > CONFIGURE_ARGS+= --disable-ident-lookups > > # - Enable Illegal Caracter "_" in the URL > > CONFIGURE_ARGS+= --enable-underscores > > # - Compile out code for ICP protocol > > CONFIGURE_ARGS+= --disable-icp > > # - Enable The use of truncate() instead of unlink() > > CONFIGURE_ARGS+= --enable-truncate > > > > recompile & install the new squid binary > > > > After that, try enabling SoftUpdates, its an special code builted in the > > kernel for the filesystems, this code will let your filesystem run as fast > > as async filesystems, but as reliable as ufs filesystems, its just > great!!! > > > > Also, mount your squid partition with the "noatime" option, you can add > this > > in "/etc/fstab" file. > > > > You have to disable memory pools in your squid.conf file, and lower > > persistent connections to 15 seconds, set all of this options to get a > real > > squid cache engine at full speed > > > > memory_pools off > > pconn_timeout 15 seconds > > half_closed_clients off > > connect_timeout 10 seconds > > quick_abort_min 1 KB > > quick_abort_max 1 KB > > quick_abort_pct 10 > > log_fqdn off > > log_mime_hdrs off > > cache_store_log none > > cache_access_log /dev/null > > ipcache_size 32768 > > cache_mem 32 MB > > > > This options has been the best ones I have found for a Dual PIII 600 1GB > > RAM, 60GB Cache Disk, it supports 9542 req/min peak reported until now for > > 6500 cable modem users. > > > > Also try not to use more than 6GB size squid cache dirs, I mean if you > have > > one 18GB disk, try using 3 lines like the following: > > > > cache_dir ufs /cache/0 6000 16 256 > > cache_dir ufs /cache/1 6000 16 256 > > cache_dir ufs /cache/2 6000 16 256 > > > > this first level & second level directory configuration has been the > fastest > > ones the default, I have used a formula floating around to calculate the > 1&2 > > levels for directories depending on the size of the partition to use, made > > some tests, used one line for 18GB filesystem, but the performance got > > degraded badly. > > > > Rebuild your kernel without needed drivers, and use the following options > in > > your kernel config file too: > > > > options "MAXDSIZ=(512*1024*1024)" #Support for a single > > proccess to use up to 512MB in RAM > > > > options "DFLDSIZ=(512*1024*1024)" #Default size of in RAM > > for a single proccess > > > > options PQ_LARGECACHE #Use all the cache > in > > the PII & PIII proccessors > > > > Add as much RAM as you can. > > > > I Hope this can help you out to get the best out of Squid & FreeBSD. > > > > Greetings... > > Ales > > > > > > > > To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-isp" in the body of the message