Re: Reserved space (WAS: How to calculate bsdlabel size)

From: Harald Schmalzbauer <h_at_schmalzbauer.de>
Date: Sat, 7 Feb 2004 15:37:08 +0100
On Saturday 07 February 2004 15:11, Erik Trulsson wrote:
> On Sat, Feb 07, 2004 at 02:17:35PM +0100, Stijn Hoop wrote:
> > On Fri, Feb 06, 2004 at 01:57:11PM -0500, Garrett Wollman wrote:
> > > <<On Fri, 6 Feb 2004 09:44:58 +0100 (CET), 
=?iso-8859-1?q?Claus=20Guttesen?= <cguttesen_at_yahoo.dk> said:
> > > > Does the algorithm(s) rely only on percentage of free
> > > > space? On a five TB (netto) filesystem eigth percent
> > > > is approx. 410 GB which seems quite alot.
> > >
> > > A good Data Structures text will prove to you that the efficiency of
> > > hashing algorithms of the sort that the UFS block allocator uses
> > > depends only on the occupancy ratio and not on the absolute number of
> > > free hash slots.
> >
> > Which translates into what wrt this question? That the 8% of free space
> > is really necessary for the UFS block allocator to function efficiently?
>
> Yes, that is exactly what it translates to.

Hm, I don't really understand why, but anyway, what performance degradation do 
I have to expect when lowering it to 2%?

-Harry

Received on Sat Feb 07 2004 - 05:37:20 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:37:42 UTC