Date: Tue, 6 Apr 2004 19:39:00 +1000 (EST) From: Andy Farkas <andyf@speednet.com.au> To: Artem Koutchine <matrix@itlegion.ru> Cc: freebsd-current@freebsd.org Subject: Re: Qustion about being "Giant-locked" Message-ID: <20040406191025.S56300@hewey.af.speednet.com.au> In-Reply-To: <004101c41bae$a9a419e0$0c00a8c0@artem> References: <C192C8912E798F4399668791C8965190674FD6@mx.hhp.local> <20040406071205.GA2819@frontfree.net> <004101c41bae$a9a419e0$0c00a8c0@artem>
next in thread | previous in thread | raw e-mail | index | archive | help
> I am trying to get your point and i really can't. I just don't see what
> your are trying to say. If you refer to 8% CPU usage then i see that it
> is only 8% because SCSI is the bottleneck. Disk just do not give data
> fast enough to load CPU high. Also, i don't think the do not give
> data fast enough only because of the Giant. I think it is mostly scsi
> and hdd speed problem in this case. Afterall md5 algorithm is not
> too havy on cpu anyway.
What I was trying to point out is that there is hardly any concurrency
with disk drivers on a SMP box.
With 4 controllers, 8 spindles, and 4 cpus to drive them, one would expect
a little bit of concurrency. Instead, the kernel spends most of its time
waiting or blocked, only one disk is accessed at a time, and the cpus stay
idle.
md5 may have been a bad example, but on a slow 200MHz cpu, it can be a
fairly intensive cpu hog.
--
:{ andyf@speednet.com.au
Andy Farkas
System Administrator
Speednet Communications
http://www.speednet.com.au/
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20040406191025.S56300>