Todd,

Do you have the no_atime_update=on option set on your filer volume(s)? If not, this can generate much activity as the filer updates the access time for each file hit. Automated testing would really whack them.

--sam

-----Original Message----- From: Todd C. Merrill [mailto:tmerrill@mathworks.com] Sent: Tuesday, March 11, 2003 2:27 PM To: toasters@mathworks.com Subject: CIFS getattr load

Instead of standing at a crossroads or a fork in the road, I'm standing in an open field, and need somebody to point me in a direction...

My filers are showing loads and loads of CIFS getattr operations, on the order of 80% of the CIFS traffic, which is 80-90% of the total NFS and CIFS traffic. Finding out why this is so, or alleviating it or reducing it so the load on the filers goes down, is my goal.

The filers are used in a software development environment, in an automated testing environment (generally no interactive usage by people). The CIFS/getattr load is most imbalanced not when a software build is happening but while testing suites are run. On the order of only 10 Windows clients (generally Win2k) can nearly cripple an F840.

So, at this point, any WAG's are welcome. Do directories with large numbers of files stress the getattr operations? Has anyone run across any applications that are very "chatty" with filers? Has anyone run into any bugs (with ONTAP or with Windows) that might show such a symptom?

Perhaps the application mix we are running is just this way, but something in my gut tells me something is amiss.

I have a sniffer available, but before I get to that detail, I was hoping to mine the wisdom of this list, to approach this vague problem armed with some good ideas.

Until next time...

The MathWorks, Inc. 508-647-7000 x7792 3 Apple Hill Drive, Natick, MA 01760-2098 508-647-7001 FAX tmerrill@mathworks.com http://www.mathworks.com ---