NetApp FAS2040 Performance Testing

Back in August we took delivery of a new NetApp FAS2040 to replace an outgoing unit. The primary function for the unit will be for backend storage for our VMware vSphere environment as well as a few MS Windows boxes (MS SQL Cluster and Citrix Provisioning Servers). Prior to us migrating production data to the unit, I spent sometime running some performance tests with IOMeter to see what it could do.

The Setup

Two Cisco MDS9124 Fiber Channel Switches
NetApp FAS2040 w/dual controllers
1 Aggregate (22 x 300GB 15K SAS drives)
HP DL380 G7 Server
Dual 4GB FC HBA’s per NetApp Controller and HP Server
ESXi 4.1 U1
Guest VM running Windows Server 2008 R2 configured with 1vCPU and 2GB RAM

Using the NetApp Virtual Storage Console I created a single 100GB LUN configured to use VMware Round Robin as the Path Selection Policy and VMW_SATP_ALUA for the Storage Array Type Plugin. The datastore was formatted with a 8MB block size. As I was creating the LUN via VSC and using Windows Server 2008 R2 on the VM I didn’t need to worry about storage alignment, but just to be sure I had a quick read of NetApp TR3747 Best Practices for File System Alignment in Virtual Environments, located HERE.

For the actual benchmark tests a 40GB VMDK was assigned to the test VM and IOMeter was used with the configuration file provided by the excellent thread on VMware Communities, “Unofficial Storage Performance Thread”. Links to this thread (now in 2 parts) are HERE and HERE.

The Performance

Each of the performance tests was completed three times and the final results the average.

 

IOPs

MBps

Avg Response Time

%CPU Utilization

Max Throughput-100% Read

11766.35

367.69

5.09

38.37

Max Throughput-50% Read

13106.30

409.57

4.57

41.99

Random-8K 70% Read

25428.99

198.66

2.31

65.90

RealLife-60% Random-65% Read

23087.59

180.37

2.55

64.29

The Conclusion

All in all I am very happy with the results. The numbers we are posting on the NetApp are better then the numbers we had on the out going unit. But as we were not bound by any performance issues in our environment do to storage and the new unit was purchase as part of a “hardware refresh”, I see this as having our cake and eating it to.

%d bloggers like this: