On Tue, 2006-04-11 at 18:13, Bart Smaalders wrote:
> 
> There is often an implicit assumption that performance is somehow
> a linear function; this is not the case for significant changes.
> There are something like 1700 packages in Solaris 11 right now.
> Assuming all packages are the same size (or they're installed
> randomly), we end up writing 16/2 = 8MB 1700 times, or 13.6 GB
> of I/O.  This is compared to the 2.7 or so GB that is Solaris
> fully installed.  Thus removing the repeated writing of the
> contents file would reduce the amount of IO needed from about
> 16GB to 2.7 GB -> probably worthwhile.

However, this analysis omits 2 important factors: (1) that the
quantity of I/O doesn't bear any direct relationship to the time
taken to perform it. While the contents file may move more bytes,
it takes significantly less time, and (2) that simple modifications
to the install process - without modifying pkgadd or the handling
of the contents file at all - can reduce the number of bytes needing
to be moved by an order of magnitude.

Looking at this, my instinct is to say that the problem is not that
the contents file is being mishandled, but that the number 1700 is
too large.

My point is not that the contents file isn't a problem. It is.
But it's impact on install times is overstated, and larger gains
are to be had elsewhere.

-- 
-Peter Tribble
L.I.S., University of Hertfordshire - http://www.herts.ac.uk/
http://www.petertribble.co.uk/ - http://ptribble.blogspot.com/



Reply via email to