From: NetComrade (netcomrade@bookexchange.net)
Date: Thu Oct 14 2004 - 12:21:47 EDT
on IO, i should've added that E4900 was running PCI cards instead of HBA,
probably a factor.
On Thu, 14 Oct 2004, NetComrade wrote:
> All,
>
> I was lucky enough to evaluate an e4900 vs e4500. Off the bet the run
> queue reduced from ~30 (pretty bad for a 10CPU e4500), to about 6 (on a
> 12cpu e4900).
>
> In addition sequential IO is running about ~30% faster (judging by
> creating an Oracle datafile), using the same hardware, same version of
> Veritas, but Sol8 instead of 7.
>
> Anyway, that was a magnificent performance boost. Today we started
> thinking that the boost is probably even better, since the run queue is
> probably per core, instead of per CPU.
>
> We decided to do a simple test, and wrote a simple script to constantly
> sit on a cpu (var=0; while var<>10 do var=var+1; var=var-1; end loop;). We
> span up 60 of those on both the e4900 and the e4500, and the 'top' command
> on e4500 showed '10 to 12 on cpu', while the e4900 showed '22-28 on
> cpu'. Now, I would think that in reality the number of processes 'on cpu'
> never got past 10 on e4500 or 24 on a dual core 12 cpu e4900, and the # is
> probably more of an estimate then a concrete number, but just wanted to
> see if anyone thinks this methodology is wrong.
> _______________________________________________
> sunmanagers mailing list
> sunmanagers@sunmanagers.org
> http://www.sunmanagers.org/mailman/listinfo/sunmanagers
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers
This archive was generated by hypermail 2.1.7 : Wed Apr 09 2008 - 23:29:34 EDT