- To: <jmd000@xxxxxxxxxxxxx>, <slug@xxxxxxxxxxx>
- Subject: RE: [SLUG] what CPU req. for squid, imap + virtual mem
- From: James_Gray@xxxxxxx
- Date: Tue, 22 Apr 2003 19:23:36 -0400
- Thread-index: AcMJGDEVhbOe1WrpT4GFlui+B0AgFQ==
- Thread-topic: [SLUG] what CPU req. for squid, imap + virtual mem
> -----Original Message-----
> From: slug-bounces@xxxxxxxxxxx [mailto:slug-bounces@xxxxxxxxxxx]On
> Behalf Of Jon
> Sent: Tuesday, 22 April 2003 8:26 PM
> To: slug@xxxxxxxxxxx
> Subject: [SLUG] what CPU req. for squid, imap + virtual mem
>
> I've got a big squid cache, and a cyrus imap spool which is
> about 110Mb.
> This is on a P3 866 + 256Mb. Max 2 users on it.
>
> Right now my mum is using a P233MMX + 256Mb with WinME and I
> didn't realise
> how slow it is! Used it the other day.
>
> My question is: will the P233 be enough for squid and cyrus?
> I like using
> squirrelmail for IMAP - it takes about 5-6 seconds to
> generate the pages for
> my Sent folder, for example. Would I be waiting a
> ridiculously long time on
> the 233?
How's this: Pentium-Pro 200, 64Mb RAM and a 4Gb UW-SCSI 10,000RPM HDD, FreeBSD *cough* 4.7. Runs an authenticating squid cache/filter (bannerfilter) for the office and feeds off a 4Mbps SDSL connection; all users have seen the full 4Mbps at the client end so there's no bottle-neck with our Squid cache. Average system load is < 0.20 and most of that is nmbd/winbind authenticating with our ActiveDirectory and the filter. Usually there are 10-15 unique users with each user running multiple requests (IE/Netscape, Win2K web updates etc). Below is a snippet from last night's system report:
These are all the processes associated with Squid authenticating with winbind. The system has been up for 6 days now (rebooted due to kernel recompile - uptime was 128 days before that....dammit!).
PID STATE TIME PROCESS
*** Windbind / Samba ***
128 Ss 1:24.58 /usr/local/sbin/nmbd
130 Is 0:20.45 /usr/local/sbin/winbindd
*** Squid+Auth+Filter ***
1946 Is 0:00.01 /usr/local/sbin/squid (wrapper)
1948 S< 5:59.67 (squid) - working program
1949 I<s 3:14.05 /usr/bin/perl - bannerfilter
1950 I<s 0:31.26 /usr/bin/perl - bannerfilter
1951 I<s 0:09.90 /usr/bin/perl - bannerfilter
1952 I<s 0:01.32 /usr/bin/perl - bannerfilter
1953 I<s 0:00.13 /usr/bin/perl - bannerfilter
1954 I<s 0:01.50 (wb_ntlmauth) (wb_ntlmauth)
1955 I<s 0:00.48 (wb_ntlmauth) (wb_ntlmauth)
1956 I<s 0:00.15 (wb_ntlmauth) (wb_ntlmauth)
1957 I<s 0:00.05 (wb_ntlmauth) (wb_ntlmauth)
1958 I<s 0:00.03 (wb_ntlmauth) (wb_ntlmauth)
1959 I<s 0:00.03 (wb_ntlmauth) (wb_ntlmauth)
1960 I<s 0:00.02 (wb_ntlmauth) (wb_ntlmauth)
1961 I<s 0:00.02 (wb_ntlmauth) (wb_ntlmauth)
1962 I<s 0:00.02 (wb_ntlmauth) (wb_ntlmauth)
1963 I<s 0:00.02 (wb_ntlmauth) (wb_ntlmauth)
1964 I<s 0:00.02 (wb_auth) (wb_auth)
1965 I<s 0:00.02 (wb_auth) (wb_auth)
1966 I<s 0:00.01 (wb_auth) (wb_auth)
1967 I<s 0:00.02 (wb_auth) (wb_auth)
1968 I<s 0:00.02 (wb_auth) (wb_auth)
1969 I<s 0:00.02 (wb_auth) (wb_auth)
1970 I<s 0:00.02 (wb_auth) (wb_auth)
1971 I<s 0:00.02 (wb_auth) (wb_auth)
1972 I<s 0:00.01 (wb_auth) (wb_auth)
1973 I<s 0:00.02 (wb_auth) (wb_auth)
1974 I<s 0:05.43 (unlinkd) (unlinkd)
So yeh, with 2 users and the system spec you describe, you should be fine :)
--James