Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 1 Sep 2003 19:14:17 +0200
From:      "mark rowlands" <mark.rowlands@mypost.se>
To:        "J.A. Terranson" <measl@mfn.org>, <freebsd-questions@freebsd.org>
Subject:   RE: Vinum stripe values table?
Message-ID:  <4789E43478F3994BB8D967C73FD9C6884EEE@EXCHSRV1.mwrwin2k.se>

next in thread | raw e-mail | index | archive | help
> J.A. Terranson
> The values to plug into vinum for stripe size has always been vague,
> other than avoiding a power of 2. =20

Where is the vagueness here?

"A good guideline for stripe size
is between 256 kB and 512 kB.  Avoid powers of
2, however: they tend to cause all superblocks
to be placed on the first subdisk."

> Is there any way to calculate an=20
> optimal value?  Or is there a table anywhere?

I did some work on this and came to the unsurprising conclusion of
"it depends"=20

The needs of a database server / web server / ftp server / user file
server are not equivalent, in addition, obviously the exact layout of
"your" vinum configuration plays a non-trivial part.=20

> Given an 8 drive raid5 array, where each drive is an identical 4096m
> drive, where do I start?

You have already started when you created an Raid 5 configuration  ;-)=20

Well, I created my vinum configuration, made some wild guesses as to
what=20
I was going to be using it for (mostly web-serving of files of the order
of
4-5MB) and then ran some tests, measuring both with bonnie and vinum's
builtin statistics using different stripe sizes and increasing numbers
of
concurrent transactions, using both bonnie and scripted file transfers
to
generate disc i/o.=20

Overall, with my config  4 disks striped raid 0 on an promise ide
controllers (real men always run Raid 0 !) an 8k stripe seemed to
function well for me in this distinctly unscientific test scenario. Real
world performance....it's ok....could it be better, probably. But it's
good enough for what this machine does.



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4789E43478F3994BB8D967C73FD9C6884EEE>