On Sat, Apr 29, 2000, Henrik Nordstrom wrote:
> adri wrote:
>
> > Modified the store API so that STCHECKOBJ will return -1 if the FS
> > is to be ignored, or a number between 0 and 1000 indicating the load.
> > diskd returns a valid load figure. aufs, coss and ufs return constant
> > load figures, which isn't very useful but its all I can think of at
> > the moment. storeDirSelectSwapDir() has been rewritten again, and
> > I'm sure its highly inefficient.
>
> I'll add a load factor to aufs later. There is a queue length which
> might be a good source for a load indicator, but I am not sure how it
> should be scaled. Or maybe I'll add some service time measurements which
> is a much better one (and can be done mostly for free).
Well, queue length for aufs/diskd isn't the most accurate load measurement,
but it'll do for now.
Adrian
Received on Sat Apr 29 2000 - 07:07:31 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:12:25 MST