Fc8 test1 kernel backported to fc7 -- qualitive impressions

[Leslie Satenstein]

I back ported the fc8 (*23*) kernel to fc7 (64bit) and noted the
following anomalies.  

When there are few processes running, performance is great.  But, here
is what I noticed and suffered with.

Start pup. (downloading 12 updates).
Started firefox and evolution only after pup started to do it's thing.  

What I experienced was very very slow startup of both evolution and
firefox(10 secs), with both the two timing out due to lack of
responsiveness.  I think that the timeout appears to be due to the pup
application hogging the cpu. 

I am guessing that pup has no timeslice max value to tell the process
dispatcher to move it to the back of the dispatch queue, or the new
process dispatcher in the *23* kernel doesn't know when it should to
give control to the other waiting processes. 

I think, by my guessing, that the timeout or response problems are due
to fc7 applications that were built for the fc7 kernel (*22*), where the
frequent clock interrupt in the kernel allowed the round robin execution
to occur equitably.  I am surmising, probably very incorrectly, that the
new kernel needs a new process dispatch algorithm which includes a max
timeslice value for that process group. It could be there, but the fc7
applications were not built to communicate that way to the kernel's
process dispatcher.   

As I indicated, my observations are qualitative. I can't substantiate
anything, so I may be completely wrong.   

My environment:  Fc7 64 bit version, home use, 1 gig ddr2 memory, dual
core processor (intel d930 -- 3gig hertz). Dual disks (one disk
exclusive for 32bit fc7, the other exclusive for 64bit fc7).

Do I like Fedora?  You answer that question.




Leslie Satenstein



-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/fedora-test-list/attachments/20070811/a81f72a1/attachment.htm>