Return to BSD News archive
Path: euryale.cc.adfa.oz.au!newshost.carno.net.au!harbinger.cc.monash.edu.au!news.rmit.EDU.AU!news.unimelb.EDU.AU!munnari.OZ.AU!news.ecn.uoknor.edu!feed1.news.erols.com!news
From: Ken Bigelow <kbigelow@www.play-hookey.com>
Newsgroups: comp.unix.bsd.freebsd.misc
Subject: Re: Apache and FreeBSD versions
Date: Wed, 25 Dec 1996 12:54:36 +0000
Organization: Erol's Internet Services
Lines: 138
Message-ID: <32C1240C.43B@www.play-hookey.com>
References: <E2FwC0.4yy@nonexistent.com> <E2MKzC.Ks7@nonexistent.com> <32BCF239.683B@www.play-hookey.com> <E2tw30.JKp@nonexistent.com> <32BD916C.4B5F@www.play-hookey.com> <E2uMxK.5z2@nonexistent.com> <32BEEE1B.7EA4@www.play-hookey.com> <E2yG1w.M52@nonexistent.com>
Reply-To: kbigelow@www.play-hookey.com
NNTP-Posting-Host: kenjb05.play-hookey.com
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
X-Mailer: Mozilla 3.0 (Win16; U)
Xref: euryale.cc.adfa.oz.au comp.unix.bsd.freebsd.misc:33125
Louis Epstein wrote:
>
> Ken Bigelow (kbigelow@www.play-hookey.com) wrote:
> : Louis Epstein wrote:
> : >
> : > Ken Bigelow (kbigelow@www.play-hookey.com) wrote:
> : > : Louis Epstein wrote:
> : [snip some older stuff...]
> :
> : > : > : > : I understand the previous discussion said 2.2,but not 2.1.6,
> : > : > : > : had been tweaked to run WWW faster,but in an ISP operation
> : > : > : > : environment stability's the key...I suppose that,with reference to the
> : > : > : > : mentioned caching program,Squid 1.0.22 would be a better bet than
> : > : > : > : 1.1.0.
> : >
> : > [just replaced by 1.1.1 in the past week...yes,1.1.0 was buggy]
> :
> : It happens everywhere. That's why I tend to sit back and wait while the
> : dust settles.
>
> They're now up to 1.1.2(not yet in the Ports Collection).
Hmmm. Evidently the dust hasn't settled yet...
>
> : >
> : > : > : My priority is also stability. I'm running Apache 1.1.1 on FreeBSD 2.1R
> : > : > : with no problems. I will leave it this way for the present, to retain
> : > : > : that stability. I won't be upgrading to either 2.1.5 or 2.1.6 on the
> : > : > : server machine, with 2.2R just around the corner. As soon as I know 2.2
> : > : > : is truly stable, I expect to upgrade to that platform, but I can't see
> : > : > : changing the platform every couple of months under my circumstances.
> : > : >
> : > : > 2.1.6 is,we have been told repeatedly,stabler than 2.2...the analogous
> : > : > release of 2.2 isn't due until summer '97.I've been running 2.1 since
> : > : > the end of '95,and if 2.1.6 is better,I figure it may be worth the
> : > : > upgrade hassle,while early 2.2.x won't be.(Just as I'd be likelier to
> : > : > go from 2.2.[max x] to 3.1 than to 3.0.)
> : > :
> : > : You're right, of course. And it's also true that 2.1.6 has some
> : > : capabilities not present in 2.1R. We'll see; I've been planning on
> : > : upgrading the hardware as well as the FreeBSD OS. I expect I'll get the
> : > : new hardware first; I've been considering going to a PCI board. But my
> : > : 33.6K dialup is still the primary bottleneck...
> : >
> : > Meaning you run a WWW server on the customer end of a dialup?
> : > (I run an ISP with a 384K line,customers presently dialing in at
> : > 28.8 though expansion will be 33.6K or ISDN or some combination).
> :
> : Yep. I paid for a 24/7 connection and a static 32-node Class C subnet;
> : this lets me have a server/gateway and still hook my little home network
> : to it. I made the server box be the gateway as well, so that packets
> : from outside don't have to fight their way over my 10B2 Ethernet as
> : well. It's not superfast, but it works and is quite stable.
>
> I may have customers in your position some time...good to know that
> it works.
>
> Have your upgrades of Apache improved performance?
Not to any noticeable extent on the output side -- I think its usage of
system resources is a bit more efficient. Here's my top output while
httpd is idle:
load averages: 0.00, 0.00, 0.00
12:38:17
32 processes: 1 running, 31 sleeping
Cpu states: 0.4% user, 0.0% nice, 0.0% system, 0.0% interrupt, 99.6%
idle
Memory: 12M Active, 1880K Inact, 6524K Wired, 7640K Cache, 3336K Free
Swap: 77M Total, 77M Free
PID USERNAME PRI NICE SIZE RES STATE TIME WCPU CPU COMMAND
498 kbigelow 28 0 428K 888K run 0:00 2.84% 0.84% top
29688 root 18 0 268K 348K sleep 0:00 0.00% 0.00% pppd
102 root 18 0 268K 336K sleep 0:38 0.00% 0.00% cron
21 root 18 0 216K 80K sleep 0:00 0.00% 0.00%
adjkerntz
456 kbigelow 10 0 672K 988K sleep 0:00 0.00% 0.00% bash
92 root 10 0 396K 468K sleep 3:36 0.00% 0.00% httpd
5508 root 10 0 464K 288K sleep 0:00 0.00% 0.00% sh
5513 root 10 0 460K 284K sleep 0:00 0.00% 0.00% sh
1 root 10 0 364K 184K sleep 0:00 0.00% 0.00% init
186 root 3 0 156K 472K sleep 0:00 0.00% 0.00% getty
187 root 3 0 156K 472K sleep 0:00 0.00% 0.00% getty
188 root 3 0 156K 472K sleep 0:00 0.00% 0.00% getty
1893 empyre 2 0 2648K 2004K sleep 4:47 0.00% 0.00% netmux
455 root 2 0 216K 572K sleep 0:00 0.00% 0.00% telnetd
5519 root 2 0 200K 552K sleep 0:00 0.00% 0.00%
popclient
Now, when something is actually happening, CPU usage of course jumps.
But even while compiling, say, Apache (which I did recently to add the
agent and referrer logfiles), the system had no trouble keeping up with
both Web page requests and the eternal cron jobs. It's also running my
duaghter's MUSH (empyre/netmux above) at a reasonable pace, according to
reports I've gotten.
>
> : > My next motherboard/CPU will be PCI,I suppose...current set is
> : > VLB at its max(486 DX4-120).
> :
> : Mine is close to that. It's currently a 486DX-50 running at full speed,
> : with a VLB 6-pack I/O card. I have one more motherboard with the AMD
> : 5x86-133 (overclockable to 160 MHz), and at the computer show last
> : Sunday I saw Western Digital Caviar drives -- the 33100 (3.1 GB) for
> : $299. They're getting cheaper by the day!
>
> You're going it without SCSI?
Yep. No SCSI, no PCI. I do have 32 Meg of RAM, and (as shown above),
about 7.5 MB of that is used as cache (very dynamic, however). I also
have 77MB swap partition on the drive, but the RAM is usually enough.
>
> : > :
> : > : As for Squid, I've seen references to both it and CERN cache in my
> : > : agent_log file, but I haven't tried it, and so have no opinion.
> : >
> : > See http://squid.nlanr.net/Squid for their claims.
> :
> : Thanks, I will. But I bet they still can't speed up that 33.6K link!
>
> Just make sure stuff is cached well on your disk.
With 32 MB RAM, I generally have plenty to allocate to cache. Between
that and 77 MB swap, I have yet to see a system complaint along this
line. (Famous last words... :-) )
--
Ken
Are you interested in |
byte-sized education | http://www.play-hookey.com
over the Internet? |