[OpenAFS] Speed of AFS, continue...

[Lars Schimmer]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

ted creedon schrieb:
| On a test server:
| 1. Play with the hdparm utility - there are some cryptic settings and
| performance tests. Use 32 bit transfers (sometimes the OS resets to PIO/16
| bits on bootup).
| 2. Invest in a PCI disk controller (tweak and benchmark the hardware
first)
| 3. CPU load of 5-6 will give you unacceptable response times - for all
| processes
| 4. Remove all unnessary software from the servers
| 5. Is the response time problem due to user complaints or just
observations
| based on a test file move?

1. I have played with hdparm ;-) So I know the UDMA mode of the servers.
2. Yes, it will be done next, but why should the onboard IDE on a
"modern" AthlonXP board be much worse than an older IDE PCI controller?
3. Yes, right, but why is load so high with AFS on it?
4. The fileserver1 has OpenAFS, Apache2 and Subversion on it. And yes,
the Apache2 is ONLY for subversion and yes, that are only 10 accounts in
subversion, and while measure that speed, Apache2 was OFF. So "plain" OS
and OpenAFS. Fileserver 2 was the same.
5. As for me being the only account on that machine, it was a ssh to
that machine while that special move of a 2 GB file. I copied the file
and tried with a second login to get some speed/load data with iftop and
top, so I experienced the lag on SSH. And no, there are no other
accounts on that machine.

6. to the other post: both machine were fresh rebootet and both are
stucked with 1 GB RAM, I assume that to be enough. And both machine has
2 GHz AthlonXP. So I assume they are NOT swapping and I assume a 2 GHz
CPU should be capable of sending 100 MBit out to the net.

OK, the database server is older and slower, but while moving one file
between two fileservers, how much load is on the database server?

So at all, I´m kinda helpless for slowness of my network.

| tedc

Bye
Lars
- --
- -----------------------------------------------------------------
Technische Universität Braunschweig, Institut für Computergraphik
Tel.: +49 531 391-2109            E-Mail: schimmer@cg.cs.tu-bs.de
PGP-Key-ID: 0xB87A0E03
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (MingW32)

iD8DBQFCM3UbVguzrLh6DgMRAmk4AJ9jCJIfVUcHha5sk76Ohu7BhKCybACgxwR3
lwVeclznTmFSnm8GE9Fop/A=
=Pwjf
-----END PGP SIGNATURE-----