Greetings. I have a X4500 with an 8TB RAIDZ datapool, currently 75% full. I have it carved up into several filesystems. I share out two of the filesystems /datapool/data4 (approx 1.5TB) and /datapool/data5 (approx 3.5TB). THe data is imagery, and the primary application on the PCs is Socetset. The clients are Windows XP Pro, and I use services for unix (SFU) to mount the nfs shares from the thumper. When a client PC accesses files from data4, they come across quickly. When the same client accesses files from data5, the transfer rate comes to a crawl, and sometimes the application times out. The only difference I can see is the size of the volume, the data is all of the same type. I could find no references for any limitations on the volume size of nfs shares or mounts. It seems inconsistent and difficult to duplicate. I plan to begin a more in-depth troubleshooting of the problem with dtrace. Has anyone seen anything like this before? Thanks. -Bob Bencze -- This message posted from opensolaris.org
I''ve had serious problems trying to get Windows to run as a NFS server with SFU. On a modern raid array I can''t get it above 4MB/s transfer rates. It''s slow enough that virtual machines running off it time out almost every time I try to boot them. Oddly it worked ok when I used an old IDE disk - I got about 20MB/s out of it then. But it''s not good when an IDE disk can outperform a raid array capable of around 500MB/s. It''s probably not directly related to the issues you''re having, but my own experience of SFU would make me want to repeat your tests with a Linux client before asuming it''s a Solaris or ZFS problem. -- This message posted from opensolaris.org
On Mon, Oct 20, 2008 at 9:29 AM, Bob Bencze <benczer at saic.com> wrote:> Greetings. > I have a X4500 with an 8TB RAIDZ datapool, currently 75% full. I have it carved up into several filesystems. I share out two of the filesystems /datapool/data4 (approx 1.5TB) and /datapool/data5 (approx 3.5TB). THe data is imagery, and the primary application on the PCs is Socetset. > The clients are Windows XP Pro, and I use services for unix (SFU) to mount the nfs shares from the thumper. When a client PC accesses files from data4, they come across quickly. When the same client accesses files from data5, the transfer rate comes to a crawl, and sometimes the application times out. > The only difference I can see is the size of the volume, the data is all of the same type. > > I could find no references for any limitations on the volume size of nfs shares or mounts. It seems inconsistent and difficult to duplicate. I plan to begin a more in-depth troubleshooting of the problem with dtrace. > > Has anyone seen anything like this before? > > Thanks. > > -Bob Bencze > -- > This message posted from opensolaris.org > _______________________________________________ > zfs-discuss mailing list > zfs-discuss at opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/zfs-discuss >SFU NFS is often slow, but tunable, here is something you might find handy to squeeze some speed out of it: http://technet.microsoft.com/en-us/library/bb463205.aspx HTH -- Brent Jones brent at servuhome.net