2014-03-19 05:50 AM
Good morning Community!
I am getting ready to implement a two node 2240 system with Clustered Data ONTAP 8.x. The implementation is a two phase process over a year timeframe. The first phase is to replace our current blade chassis config running VMware with Netapp NFS providing the storage for the datastores. Phase II would be able to provide some CIFS shares for some windows servers at the remote location. We have 24 drives in this configuration - 3 drives for node1 aggr0, 3 drives for node2 aggr0, 2 drives for spares, and the remaining 16 drives for data aggregate. To ensure we get maximum performance, I have decided in one large aggregate served from one node. So I will lose the performance of utilizing two nodes, but I will have a good spindle count and a failover node if there is a problem. The recent training Clustered ONtap training class I attended suggested I create a vserver for NFS and a vserver for CIFS. I was interested in opinions on this configuration. The other topic I was interested in was the network configuration. To ensure I have adequate bandwidth for the NFS traffic from VMware, I had planned on creating an interface group with e0a/b/c/d (1gb ports) and then creating vlans on top of this interface group for nfs traffic, cifs traffic, and mgmt. traffic -- thoughts/concerns, ideas. Also I would be interested in any insight from anyone that has deployed VMware with NFS via Netapp. Thanks,
2014-03-20 08:14 AM
Not sure if this will be a problem for you but make sure your VMDKs are aligned. Misalignment seems to be the biggest performance issue on NFS datastores on FAS, especially since you'll be utilizing the power of only one node.