On Tue, 2006-04-11 at 11:21, Peter Tribble wrote:
> This tells me that the pkgadd process itself
> is inefficient or has some other overhead that
> needs to be identified.

So I ask myself the obvious question:

How long does it take to actually write Solaris
to the disk?

It is, after all, 3.5G of data in 130k files or
directories.

By comparison, /opt/sfw (1.4G of data in 80k files)
takes 6 minutes to untar (5 minutes if I enable
the disk write cache). It's probably fairer to
scale on the number of files than the volume of
data, but even so that means that writing out the
real data is taking 10 minutes or more.

(The problem is that Solaris has an average file
size of 30k or so. Writing such small files - at
least with ufs - isn't terribly efficient. For
my disk 30k at 200 io/s makes 6M/s, so that 3.5G
is expected to take 600s or 10 minutes, which
agrees with the estimate above. At least the
contents file is splurged efficiently, in 1M
chunks, allowing the drive to go flat out - at
40M/s the 6.5G of contents file writes would
only take 3 minutes, which is what I'm seeing.)

-- 
-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