Re: [squid-users] cache_dir ... plz help

From: Henrik Nordstrom <hno@dont-contact.us>
Date: Thu, 01 Nov 2001 20:06:41 +0100

Right. I must have been seriously sleep deprevated when I wrote the
message below.. but I wasn't that far off, only 2^2.

The correct formula reads:

L1 = cache_dir_size * 2 * 1024 / (average_object_size * L2 * L2)

which when simplified equals to

L1 = cache_dir_size / 416

Regards
Henrik

Scott Anderson wrote:
>
> That does not explain the number 416. If i take 2 * 13 * 256 * 256 this
> equals 1703936 which is a far cry from 416
>
> Henrik Nordstrom wrote:
> >
> > timo.dudziak@student.uni-siegen.de wrote:
> >
> > > cache_dir [ufs | asyncufs] dir MB L1 L2
> > > perfect for ext2:
> > > L2 := 256
> > > L1 := ceil( MB / 416 )
> > > so why / 416 ?
> >
> > Because someone simplified the calcuation for you by eleminating all
> > constants.
> >
> > L2 = 256 (there is generally no reason to change this on
> > ext2/ext3/ufs/whatever with a similar directory structure)
> >
> > L1 = ceil( MB * 1024 / (2 * average_object_size * L2 * L2))
> >
> > MB = size of the cache in MB
> > 2 = a safety margin
> > 1024 = conversion factor from MB to KB
> > average_object_size = 13 KB (farily standard)
> >
> > For reiserfs there MAY be a slight benefit in increasing L2 somewhat,
> > but not too much. There is only marginal differences in tuning this
> > parameter.
> >
> > Regards
> > Henrik Nordström
> > Squid Hacker
>
> --
> It is by caffeine alone I set my mind in motion,
> It is by the Beans of Java that thoughts acquire speed,
> The hands acquire shaking, the shaking becomes a warning,
> It is by caffeine alone I set my mind in motion.
>
> Scott Anderson
Received on Thu Nov 01 2001 - 12:07:33 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:03:49 MST