[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] >NFS performance issue on Xen virtual machine
It has been decades since I've used NFS and have never used Xen with NFS, but I distinctly recall that CPU latency can be a huge problem with performance. The one instance I recall was that I ran 4 cpu-intensive processes on a 4-cpu nfs server, and the context switching time of the kernel caused all clients (dozens to hundreds) to become significantly (10x?) slower. Running just 1 cpu-intensive process still noticeably degraded the server's performance. I would suggest making sure that all the CPUs are idle, perhaps test with more cores? (i.e. more than 1) In general NFS is very very latency sensitive. TCP might help with throughput (reading one large file) as it's been known to help with far away (25 ms) servers, but for transactional cost (i.e. open/write/close/stat) lots of files, it shouldn't. - Kevin
|
Lists.xenproject.org is hosted with RackSpace, monitoring our |