Install Guide

25 Best Practices for Sharing an iSCSI SAN Infrastructure with Dell PS Series and SC Series Storage using VMware vSphere
Hosts | 2015-A-BP-INF
6 Test methodology
Note: The test methodology used in this paper is only valid in the presented topology, components, and test
cases. The results may not be directly applicable to environments with other components or variables.
Additionally, the updated versions of vSphere, PS Series firmware, and SC Series SCOS show no
indications of presenting different performance than that of previous versions. As such, the performance
testing reflects the previous version of this document.
Key testing factors discussed in this section include:
 The test cases assess the impact from the operational and performance perspective of the shared
host, host NICs, and shared Ethernet SAN infrastructure using the iSCSI protocol.
 The main objective was to ensure that the VMware vSphere hosts with both shared and dedicated
NICs and shared Ethernet switches would sustain I/O workloads on Dell PS Series storage when Dell
SC Series storage is added.
 SAN component failures were induced to confirm that the solution is highly available, and that a
failure on one type of storage did not impact the other.
 The tests focused on validating the capability of the hosts and switches to handle both PS Series and
SC Series storage with respect to network adapters on the host, switch hardware, buffer settings on
the switch, Ethernet flow control, and resiliency of the storage network.
A baseline test running a specific I/O workload was performed with PS Series storage in a standalone
environment. The standalone environment consisted of the VMware vSphere hosts accessing only PS Series
arrays. The test was then repeated in both dedicated and shared coexistence environments (including
PS Series and SC Series storage) with both storage types sharing the Ethernet iSCSI SAN to ensure that the
data could be compared directly without introducing variables beyond those specified in the test plan. A high
availability test for link failures, switch failure, and other factors was carried out to prove solution robustness.
In order to determine the relative performance of each SAN design, the performance tool vdbench was used
to capture throughput values at four distinct I/O workloads:
 8 KB transfer size, random I/O, 70% read
 64 KB transfer size, random I/O, 70% read
 256 KB transfer size, sequential I/O, 100% read
 256 KB transfer size, sequential I/O, 100% write
Note: Information about vdbench is available at http://sourceforge.net/projects/vdbench/.
The above I/O workloads were first run for one hour each in a series with vSphere host servers accessing
PS Series volumes only in the iSCSI SAN network to establish a baseline. Next, the same I/O workloads were
run for one hour each with each of the vSphere hosts accessing either PS Series or SC Series volumes
sharing the same iSCSI SAN network. Finally, the same I/O workloads were run for one hour each with both
of the vSphere hosts accessing both PS Series and SC Series volumes using the same host NICs with both
storage platforms sharing the same iSCSI SAN network. The results for both shared and dedicated tests were
compared against the baseline for any change in performance.