exudnw@exud1.ericsson.se (Dave Williams) (05/21/91)
I have a few questions regarding client NFS performance. Please forgive my ignorance, but I can't seem to find the RFM to answer these questions. 1) Is there any way for a client to measure the response time for a server running normal, everyday tasks? I know I can run nhfsstones all day long, but I want to know how my servers are responding to real loads in real time. My typical client is a diskless SLC sharing a 4/490 with about 25 others on a subnet. 2) Do NFS writes to a server from a diskless client show up as NFS writes (at the server, as seen by nfswatch 3.0) regardless of their type (swap, tmpfs, etc...) or are some of these actions "below" the file systems? 3) What tools are available to test NFS and general filesystem throughput? I've got iozone and nhfsstone, are there others? What/where are they? I'm trying to find out what the limiting factors are for performance in our environment, and how to fix these (Auspex, Omni/Legato, etc...) any thoughts on this subject would be welcome. -- = exudnw@exurchn1.ericsson.se || dnw@ponder.csci.unt.edu (214)907-7928 = = David Williams = = Ericsson Network Systems = = Richardson, TX 75081 These opinions are mine - not my employers. =
barroso@priam.usc.edu (Luiz Andre Barroso) (05/21/91)
exudnw@exud1.ericsson.se (Dave Williams) writes: >I have a few questions regarding client NFS performance. Please forgive >my ignorance, but I can't seem to find the RFM to answer these questions. > > ... I'd like to add to your question another item: Is there any way for a system administrator to collect traces of file system accesses in a network file system like NFS? I mean, in the level of "opens", "reads" and "writes" to a given file. Thanks, Luiz A. Barroso -- EE-Systems, USC, LA, CA, USA.