Re: Disk I/O

[Christopher Reid Palmer <reid@pconline.com>]

Randy,

> Forward at will.

Did. The TCLUG peeps will soon start saying how this is getting boring,
then I'll stop. :) (Forwarding, that is; I won't get bored for a while
yet.)

> Do you have an AGP motherboard?  Is the AGP on a PCI brdige, or a separate
> channel?  If it is effectively on the PCI bus, it could screw peak transfer
> rate.

Yep, I have an AGP mb and my video adaptor is AGP. I don't know if it is
on the PCI bus or seperate. Given that is was a bargain board when I got
it a year and a half ago, I imagine it's probably shared.

I guess that if everything were like AGP--with direct access to
memory--things would be happy like in a Sun or SGI workstation.
(Accelerated SCSI Port, Accelerated Network Port, ...) What are the
reasons such architectures don't exist in the PC world? Is it more
expensive to build or design than a traditional PC bus architecture? Is it
just not worth it to throw-away-mentality PC makers (who expects a new
Compaq machine to still be here in six months)?


> >Netscape (thanks for the link Carl!) and checking my email. It's useless
> >to measure the performance of a machine sitting still, because machines
> >don't sit still.
> 
> Its also stupid to measure the performance of a machines subsystem when
> you're pre-empting it.  You need to characterize the nodes, their
> interrconnects, and then the full system load behaviour.  Otherwise
> everything if for naught, and senseless griping.

In a single-tasking batch-y environment, yes. I don't live in such an 
environment. I wanted to run a benchmark in the same environment I run all
my other tasks. My processes are constantly pre-empting each other:
Armadillo compiles, Netscape loads, I check my mail, damn that GIMP filter
is taking forever, someone hits the wrong mouse button and pastes 2KB of
personal email into their IRC window... A benchmark is meaningless in an
ideal-scenario vacuum. What I wanted to find out was, while I'm doing my
normal things, what kind of disk I/O can a big background process like a
compile expect to get?


> >What worries me is the huge CPU eating rates. The Commentary doc that
> >comes with Bonnie suggest that's a fault of Unix filesystems.
> 
> A combination of lazy writes, resource management & allocation overheads,
> context switching, event handling (interrupts), etc.  Raw partitions make
> benchmarks look pretty.  Thats about it.  How long of scatter/gather lists
> can be queued to the adapter?

It's a Mylex/BusLogic FlashPoint DL, BT-932, and its manual doesn't seem
to say anything about scatter/gather lists.


--
Christopher Reid Palmer : www.innerfireworks.com