I am running netapp on tap 18.104.22.168 in a 2008 R2 environment and am getting terrible restore performance. I average about 10mb/s and I am not quite sure why. Non-NDMP restores locally to our Networker server hit about 50mb/sec. Our network is 1gb all the way through with full duplex enabled. Anyone have any ideas?
I have a similar performance issue with restores from my Celerra-based NDMP clients. Any sort of "read" operation on my Celerras seem to under perform. The restores coming from the NAS boxes are generally small, so that hasn't been a show stopper. But any clone/stage operations see the same weak performance... generally hovering around 10-25 mb/sec. My first thought was maybe a block size mismatch between the virtual and physical devices used for staging/cloning, but that wouldn't explain the same performance on the restores. Hopefully someone has some good input...
JasonLee thanks for the response. Tomorrow I think what I am going to do is bring up a networker server in the same subnet as the netapp filer and see how the restores go. I shouldn't have to do this but it seems like either my networker isnt able to push the data out fast enough or my netapp filer isnt able to take in the data fast enough but that's where I am right now. I hope there are others in the community who can provide some input.
Emc mentioned to me before that cloning throughput is much lower than your backups because networker isn't multiplexing the data. Far as I know there aren't any plans to fix this. I get between 25-30mbps when cloning my backup data offsite.
Restores probably work the same way as the clones. /shrug
I apologize I am not familiar with the terms as this is a new implementation but I know I am not directly attached to a filer and I am not backing up to tape, I am backing up to disk so I must be using Networker DSA.
Could you check option ndmpd.tcpnodelay.enable on filer. If it is set to "false", could you try changing it to "true" and repeat restore. There are chances that it will help. I have seen exactly the same issue on multiple installations, but had not yet got possibility to systematically debug it.
To change option - log in to filer using command line (telnet by default) and use
options ndmpd.tcpnodelay.enable true
I will check this out today and get back to you. At least let you know what the setting is. I can change the setting afterwards and test the backup. Thanks for the post and I hope this answers my question.
This seems like it will help both the backups and restores.