Search squid archive

Re: Hit-ratio graph: can anyone explain what we see here?

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On 25/02/2013 9:08 p.m., Coenraad Loubser wrote:
Hi All

I'm very curious so as to everyone's casual explanations for what
could cause the up-and-down pattern visible in this hit-ratio graph:
http://imm.io/XkQ4

I have my own ideas but will reserve them for the time being so as to
allow you to share an unbiased opinion first. :-)

* Interaction between browser caches and Squid cache leads to funky diagrams. All objects which are cacheable by Squid are also cacheable directly on the client machines. The objects are usually cached at both places. So popular objects which stay popular can cause cycles of HIT/MISS as clients caches expire and re-fetch them. The first client will cause a new copy to be stored in Squid which will cause a peak of HIT traffic as all other clients expire and HIT on the new Squid opy. This trails off as the other normal MISS traffic flows in afterwards.

* You can also get cycles if your cache is too small to store all the popular "hot" objects your clients are wanting. Squid garbage collection coming around every 1-2 hours purges space to make way for new traffic and can unwittingly erases some of those HIT objects, causing a drop in ratio at your MISS rate increases.

* If you are running a reverse-proxy spiders can also hit you in waves as they all apply similar delay between requests, and they tend to raise the MISS rate too.


Lots of possibilities. But ...

... what Squid version? (squid -v version line please)

... and what are your ideas?

Amos


[Index of Archives]     [Linux Audio Users]     [Samba]     [Big List of Linux Books]     [Linux USB]     [Yosemite News]

  Powered by Linux