home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!math.fu-berlin.de!mailgzrz.TU-Berlin.DE!mailszrz.zrz.tu-berlin.de!woks4000
- From: woks4000@mailszrz.zrz.tu-berlin.de (Wolfgang Ksoll)
- Newsgroups: comp.unix.aix
- Subject: Tuning NFS-Server?
- Date: 3 Jan 1993 14:59:38 GMT
- Organization: CompuNet Berlin
- Lines: 33
- Distribution: world
- Message-ID: <1i6v0qINNkf9@mailgzrz.TU-Berlin.DE>
- NNTP-Posting-Host: mailszrz.zrz.tu-berlin.de
-
- #include <season.greetings.h>
-
- Hi Folks,
- let's assume the following situation: I have a 320H acting as /usr- and
- /home-server for some 5-10 clients (small machines 320 or 220).
- The OS-level is AIX 3.2.2.U`date`
-
- If heavy metal compiling occurs on all clients I can observe, that the
- server goes to 100% sys load (I use monitor 1.06). The more the clients
- push NFS-load the more response time goes to infinity.
-
- Before I replace the 320H with a 990 and the Ethernet with FDDI or SOCC
- I would like to tune the NFS-perfomance of the 320H a little bit.
- So I have three questions:
-
- 1.) Can I disable UDP-checksum calculation?
- I have no reference whether it is enabled at all, but I have heard from
- other vendors that this would be one possibility to increase NFS-throughput.
-
- 2.) Can I enable asynchronous write?
- I remember that someone told me that in spring. But I found no hints
- in info or smit. I would like to have every write-call return
- immediatly regardless of what is happening really down the pipeline.
-
- 3.) What else can I do?
- (yes, I know Presto and Auspex, but I want to tune the 320H ;-)
-
- Thank you all,
-
-
- Wolfgang Ksoll woks4000@mailszrz.zrz.tu-berlin.de
- CompuNet Berlin GmbH, Mariendorfer Damm 1-3, W-1000 Berlin 42, Germany
- Phone: +49 (030) 707 85 - 123 FAX: - 130
-