Re: Disk layout for GC

[Arie Shoshani <arie@math.lbl.gov>]

> I have a configuration question.
> 
> What are the Unix disk requirements for GC ?
> (The disk space I am talking about is where the files get placed after 
> being pulled out of HPSS.)
> 
> Does the GC need a fixed allocation of space (i.e., it's own disk
> partition) or can it share disk space with other processes ?
> 
> 
> Thanks
> 
> Shigeki

As of now, we'll need a separate disk partition for the Storage Manager, 
because we are assuming that a fixed allocation is available to us.  Any 
tests we wish to perform are based on assuming that a certain amount of 
cache is available.

What is the largest partition that can be allocated to GC experiments?

For the future, we are considering the effect of sharing the cache with 
ad-hoc objectivity reads (derefs).

Arie.