Network and Storage Protocols

Increase CIFS Performance


Hello all, a client of ours is running NetApp FAS3020 Cluster with v7.2.6.1P2. The have 3 DS14-Mk2-AT shelves with 636GB SATA drives, i think they are 700GB drives but show up as 636GB. There are 2 aggregates setup on the sata shelves with 14 drives in the Raid Groups. There are 2 GIGe ports setup in a single mode vif for cifs with default MTU of 1500. The cifs is setup on one of the sata aggregates. I see spikes on the cifs protocol on the filers, sometimes reaching 2100 OPS/Sec. at the time of the high cifs ops, the CPU Usage is at 14.3% and Disk Usage is 11.1%. End users complain that sometimes it takes a while for their file copy or other cifs operations.

So  my questions is, would it help if I enable jumbo frames on the switch and netapp filer cifs vif or change the vif from single to either static or dynamic multi vif? Or what else can I do to increase this cifs performance? We are going to add another shelf of same kind of sata drives to the system and increate the aggregate to get more space but hoping to also increase the spindles in the aggregate. Any suggestions to help in this matter would be greate appreciated. thank you in advance.




Your best bet is to open a case with NetApp and let them help you get perfstat data to try and determine where the bottleneck lies. There's a good chance it's not the storage but that can at least rule that out.

-- Adam Fox


Complete agreement with Adam's recommendation as there can be a lot of places to look here. Various items you might end up poking around at are....

  • Reallocation potential on the volumes in question (search NOW for "reallocate")
  • Network traffic on the NetApp NICs (can see either from Operations Advisor/Performance Advisor or from the switch side if you have something like Cacti in place)
  • Operations Manager or Performance Advisor (especially for the real time data).