Andrew Reid <rei...@bellatlantic.net> wrote: > There is a performance-tuning section in the NFS Howto, > with several tips, including simple tests for measuring performance. > <http://nfs.sourceforge.net/nfs-howto/ar01s05.html>
Thanks for the reminder. Somehow I'd missed that section of the Howto. > My own experience with NFS on a server with several tens of > clients is that the biggest determiner of performance is the type > of file operations you're doing. A Java compile on a tree mounted via NFS is the biggest hit (IO bound, not compute bound!). There's the usual edit stuff associated with a perl/html website, but that's one (small-ish) file at a time and the delay is not so noticeable. Oh, and a development installation of ORACLE RAC is using another NFS partition for its stuff, but again this isn't very heavy usage. One thing that I discovered from the NFS Howto is that our NFS server was running only a single nfsd thread. That, despite the /etc/default/nfs-kernel-server file having a definition of RPCNFSDCOUNT=8. I don't see why the 8 is being ignored in favour of 1, but that's for another day. Thanks, Chris -- To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
