Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/18/84; site wdl1.UUCP
Path: utzoo!watmath!clyde!cbosgd!ihnp4!alberta!ubc-vision!uw-beaver!tektronix!hplabs!hpda!fortune!wdl1!jbn
From: jbn@wdl1.UUCP
Newsgroups: net.unix-wizards
Subject: SUN NFS tuning query
Message-ID: <709@wdl1.UUCP>
Date: Fri, 20-Sep-85 21:29:13 EDT
Article-I.D.: wdl1.709
Posted: Fri Sep 20 21:29:13 1985
Date-Received: Tue, 24-Sep-85 03:41:59 EDT
Sender: notes@wdl1.UUCP
Organization: Ford Aerospace, Western Development Laboratories
Lines: 16
Nf-ID: #N:wdl1:64000015:000:747
Nf-From: wdl1!jbn    Sep 20 15:47:00 1985


     Is there any way to measure the buffer cache hit rate on a SUN file
server, lacking source?  Also, does SUN's kernel have hashed buffer lookup,
or will performance degrade if you have huge numbers of buffers?  Advice
on tuning NFS systems would be appreciated. 

      Incidentally, we've discovered that the most common file server operation
here is "lookup" (of a pathname), not read or write.  This seems strange,
until you realize that reads and writes may be satisfied in the station
caches, but lookups cannot be; clients must go back to the server for every
open operation; if they didn't, a client could open a file after another
client had deleted it.  An interesting side effect of SUN's stateless 
implementation.

					John Nagle


Brought to you by Super Global Mega Corp .com