Re: Successfuly Ported/Compiled Squid on NT

-----BEGIN PGP SIGNED MESSAGE-----

Henrik Nordstrom writes:

| With these two paches b19 should configure, build and run without any
| changes or configure time options. I have very limited testing abilities
| in my environment and would preciate if someone could do some additional
| tests before we publically announce than Squid is ready for beta testing
| on the NT platform.

I've built and done some testing of b19 on an NT server box here,
which one of our systems bods (cheers, Dave!) kindly let me play with
- - people should be able to get cache manager results by pointing their
1.2 cachemgr client in the general direction of ccdmt-server.lut.ac.uk
[131.231.133.150], port 3128. Needless to say, this is just a quick
hack to (dis)prove Squid on NT using the gnu-win32 stuff, and the
server *will go away* ! Don't say you haven't been warned!!

Unfortunately, I'd done this before I saw your mail with those two
patches. However, things seem to be holding up pretty well...

Building with the Cygnus b19.1 release, vanilla 1.2 b19, and :-

  ./configure --enable-gnuregex

I do get a problem with the shar archive of icons for Squid - no
uudecode with Cygnus' distribution of essential Unix utilities! Do the
icons need to be shar'd up for any particular reason ?

I'm also slightly confused as to whether Squid needs the error
messages and icons to be present in its default configuration - they
don't seem to be installed when you do a "make install", but Squid
seems to assume that they'll be in /usr/local/squid/etc ?

OK, so on to the NT bit...

I actually didn't run Squid out of "/usr/local/squid", so can't
comment on that bit of the installation. Will have a go at this
tomorrow or early next week.

I ran up eight tcp-banger and eight udp-banger processes, all bashing
away at the same machine at the same time, each with a different
collection of URLs taken from a recent day's JANET cache traffic. All
told, there are just over two million URLs in this test set.

If the cache manager stats are to be believed, at one point Dave's PC
(a lowly 100MHz Pentium with EIDE disk and 20MB RAM, 1GB cache) was
successfully handling sixty ICP requests per second (it was reported
as 30000 per minute by the cache manager) and after four hours has
managed to get through nearly 1.5 million of the ICP requests. Phew!

The proxy HTTP requests seem to be being serviced quite slowly -
possibly due to the overhead involved in terms of CPU usage (I've been
seeing between 30% and 50% CPU usage most of the time) and the effort
involved in writing to the EIDE disk. I've probably slowed things
down a bit more by being a pain and running "tail -f" on the cache and
access logs, though :-)

More importantly, though - it's still up!

So, kudos all round :-))

Ciao!

Martin

-----BEGIN PGP SIGNATURE-----
Version: 2.6.3i
Charset: noconv

iQCVAwUBNTZWstZdpXZXTSjhAQG2GQP/VFqqsx4otW8zKHiaRfT4TwBmZOE07Zuy
LOcU05Vy2dVqoCTFAltlWGmoLGUCDA5IEX5ubgh0mxdeuWONYm0mvjBurNc0Dknc
wTlv5clZ16X1GUcALfymFf9A209NsV3Vr43TlBavYRX+rzsvzt4a5YSwIqhe/fbw
I57y43iDEWw=
=lXuh
-----END PGP SIGNATURE-----
Received on Tue Jul 29 2003 - 13:15:47 MDT