Clamd Verses ClamAVModule timing

[Res]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Sun, 3 Jun 2007, Rick Cooper wrote:

> All of the tests I did were with attachments, but not large. Maybe I was
> just curious as to the differences and the percentages were pretty even, of
> course load would be a factor as I didn't do it with a isolated box. It
> seemed like the bigger the batch the more clamd had over clamavmodule but
> one would assume that to be true as clamavmodule presents the files one at a
> time where as the current clamd code hands the entire batch at once.

One clam per batch would be very fast compared to other, it was the 
one clam process per msg that saw us dump qmailscanner, load was making 
the machine unusuable, till I gambled and installed mailscanner on them.

I would be interested in seeing the results from clamd against f-prot, we 
found with clam in the 'old way' kept the machines loaded at around 9,
with module around 5 and with f-prot the load was under 1, but because
of f-prots complex and crazy licensing (I'm not paying for a million user 
licence for a machine that might have that many destination addresses or 
might not, eg: secondary MX's, FSI told us if we think itll do that many 
then thats what we need to buy) so we still use clam on 99% of the servers
and until FSI wake up and change the policy, we wont be purchasing any 
more, nor if I can help it will we renew, which is why I'm following your 
changes very very closely.


- -- 
Cheers
Res
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)

iD8DBQFGY1WksWhAmSIQh7MRAmsxAJ9gIPLdZV6n2UrOCTFEudufUDx5gACfZYGV
3rqI1asUV/mRTVbJedh8xks=
=5Pcc
-----END PGP SIGNATURE-----