Subject: Re: CCD performance tuning
To: NetBSD Users <>
From: Uwe Lienig <>
List: netbsd-users
Date: 01/22/2002 18:02:21
Excuse me, forgot to send this message to the list. Only send a reply to Aaron.

"Aaron J. Grier" wrote:
> On Tue, Jan 22, 2002 at 09:01:46AM +0100, Uwe Lienig wrote:
> > right at the beginning of the year I've pushed a question onto the
> > netbsd-pmax-list. Since nobody replied to it, I reposted my question.
> > Since the question isn't pmax-specific I was given the advise to post
> > this question to this list instead of the platform list - so here it
> > goes.
> what is the question?  I see a nifty shell script for testing speed of

yes nifty, but I wanted to know how the interleave will affect real, sys and user time. OTOH this
was a simple burn in for the disks without the necessity of running bonnie. May be it was not a very
clever approach, but this one runs without any additional tools.

> ccd arrays and results of running your test, but I see no direct
> question posed.

Yes, you're right.

My question is: what would the best parameter settings be? I don't have the choice to change
(replace them by better hardware - although I do have spare disks) any of the parts involved in the

The only parameter to be changed is the interleave. It should give some information whether the disk
writes are arranged in a manner that every disk gets somehow the same data stream increasing the
total perfromance of the ccd. The results give different values for real, user and system time.
Sorted by the system time the best value for interleave would be 432. But really ? I ran the test
with an interleave increase of 1 in the range of 64 to 128. But this test didn't made me feel

Reading ccd(4) one would conclude, that the interleave should be the size of a track. That is
somewhat funny since SCSI disks ( and their other counterparts e.g. (E)IDE ) don't have a fixed
track size. And even the involved RZ58 has 85 sectors per track. But setting interleave to that
value does not yield to the lowest real execution time.

Another question evolves, if the user and sys times are compared. Sometimes the real time is low.
Nevertheless sys time is much higher than in other cases. Changing interleave from 448 to 456
increases sys time by roughly 30%.

> if you are wondering whether or not your results are good, I would
> suggest that you use additional metrics to dd.  raw I/O is only one part
> of disk performance; you have ignored seek time, for instance.  there
> are tools in pkgsrc (bonnie and iozone come to mind) which can be used
> to get a better feel for overall performance.

May be overall performance is better measured by bonnie and the like. But will this nifty shell
script give some reasonable advise as well?

But I see, that I should use those tools as well. But this will impose a long test. To avoid a very
long test it seems to me that dropping some interleave numbers would be a good point to shorten
further test.

> --
>   Aaron J. Grier | "Not your ordinary poofy goof." |
>           "Making people dance so hard their pants almost fall
>                  off is kind of fun."  -- David Evans

| Uwe Lienig                                                                               |
| Forschungsinstitut Fahrzeugtechnik          Vehicle Research Institute                   |
| Fachbereich Maschinenbau/Verfahrenstechnik  Department of Mechanical/Process Engineering |
| Hochschule fuer Technik und Wirtschaft      University of Applied Sciences               |
|                                                  |
|                                                         |
| phone: (+49 351) 462 2780                   fax:     (+49 351) 462 3476                  |
| parcels: Gutzkowstr. 22                     letters: PF 12 07 01                         |