Re: Most than 1 million Req per day

From: George Michaelson <ggm@dont-contact.us>
Date: Tue, 13 Aug 1996 08:33:09 +1000

  For mine, I use the TCP hit rate as a approx hit rate.
  From my understand, every ICP hit will be translated
  (eventually) to a TCP hit. Because ICP doesn't transfer
  the data. So, ultimately, all hit should be judged
  from TCP hit.
  
  Correct me if I am wrong.

Squid-1..04 (?) introduced ICP piggybacking requests if the data will
fit inside some UDP/ICP limit. This means that no, not all ICP will
become TCP and thus you are undercounting "hits" on your cache.

Simply summing ICP and TCP overcounts hits for the same reason: not *all*
ICP will have data.
  
I'm running the cache-stats scripts off the nlanr pages. If somebody
wants to suggest which fields we should be summing on this, I am very
agreeable to us floating hitrates (I'd love to know relative performance
issues especially in the context of memory/disk cache sizes) but
I'm not comfortable with the current single-digit number: I need to know
a basis for comparison first!

-George

--
George Michaelson         |  connect.com.au pty/ltd
Email: ggm@connect.com.au |  c/o AAPT,
Phone: +61 7 3834 9976    |  level 8, the Riverside Centre,
  Fax: +61 7 3834 9908    |  123 Eagle St, Brisbane QLD 4000
Received on Mon Aug 12 1996 - 15:35:02 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:32:47 MST