On Thu 02 Oct 2008 at 05:11PM, [EMAIL PROTECTED] wrote:
> Folks,
> I finally have a fix for 3505. This reduces the amount of time we spend
> processing manifests and actions before and during the evaluate stage.
>
> In the performance tests that I ran, this took 10 seconds off the
> evaluate time, reduced user cpu time by 8-10 seconds correspondingly.
> It also reduced our wall-clock runtime by 20 seconds. The caching
It helps to know "10 seconds out of how many?" Can you post some
before-and-after comparisons of e.g. 'install -nv' and
'image-update -nv' using ptime?
> doesn't appear to have any negative impact upon the RSS of the pkg
> client.
That surprises me a little. Is it possible that in the old code
we weren't discarding manifests once we turned them into actions?
All of the manifests on my system look to add up to about:
$ pkg contents -m | wc
285781 2304386 52011853
So I would expect us to need an extra 52MB on a system with
redistributable installed?
-dp
--
Daniel Price - Solaris Kernel Engineering - [EMAIL PROTECTED] - blogs.sun.com/dp
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
