--Signature_Thu__20_Oct_2005_06_53_27_-0400_ovKvKq_fkLs7oFUl
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: quoted-printable

Hello,

> > I have a 300MHz PowerPC box here ( running AIX ) which doesn't get
> > anywhere near breaking into sweat with 5 clients building NetBSD
> > releases at the same time entirely via NFS so I guess you won't even
> > need a modern PC to serve 20-30 diskless clients.
>=20
> I wonder how to find out the sweat level of such a machine,
> for both disk and network I/O.=20

As far as I can tell it's not so much the amount of data it's supposed
to move but the number of requests to process and the number of
fragments to read to / write from filesystems.

> Are the drops in "netstat -d -w 1" and the MB/s in "iostat -x -w 1"
> the numbers to monitor? With the latter one compared to the max.
> throughput that the disk/bus can do?

Comparing a disk's momentary troughput to the buses max doesn't make
that much sense, a disk won't necessarily be able to reach it and even
if it does you'll rarely reach the max in real life unless you're
accessing huge, unfragmented files, probably sequentially.
Watching transactions per second and bus congestion seems to make more
sense, disks can be 100% busy ( and thus make the machine sweat ) when
transferring only a couple of kB per second. Kernel time seems to be a
useful indication too.

have fun
Michael

--Signature_Thu__20_Oct_2005_06_53_27_-0400_ovKvKq_fkLs7oFUl
Content-Type: application/pgp-signature

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.0 (NetBSD)

iQEVAwUBQ1d3K8pnzkX8Yg2nAQKYhQf/RLV1EDC6FqRgaDGOeHIG+jP2EcMdJ40h
ijev0UXZpr5pK9j5FSZGZjM/8VwJ6/fYk925HYFQk+LSfrgwOywWqQX8oxIDQjwj
IoJZpInwnri5pMu8pztYK0Lij+YXN57QQtGKwRlcL9EqzS7MIDzSnlcJGg9U8W5c
V/zYLX8lhbqw3EDmWwmgR/dAc3/iCMnoNJWjLpUpxhybGtVcBDCCMopJe9uXNuWe
brSGQqKDuB4OOrpuod8VQEFxmCxa1Yee3jc9nUPzIQfFCrr+p3QGPcY/RQnbmbmp
DDOmsoZ8GbIfchei5DOU8Fx1Day+4kU29xWHg+gsRDyQ+Odu/v7LbA==
=Dta8
-----END PGP SIGNATURE-----

--Signature_Thu__20_Oct_2005_06_53_27_-0400_ovKvKq_fkLs7oFUl--