Dell PowerVault MD3200i - Page 7

Dell PowerVault MD3200i/MD3600i Configuration Guide for VMware ESX4.1 Server Software
June 4, 2012 Page 7
ARCHITECTURAL SETUP
As a best practice, Dell recommends using a separate Gigabit Ethernet network switch to handle
iSCSI storage traffic, see PowerVault MD32x0i IP SAN Best Practices
. An IP SAN consists of one
or more hosts, connected to one or more storage arrays through an IP network, utilizing at least
one switch in the network infrastructure. Each switch in the configuration has a path to the
MD3200i via duplex quad-port controllers.
There are many ways to implement an IP SAN based on need, available resources and intended
application. Some general rules can be used when designing your IP SAN. For example, in order
to maximize the throughput of your storage arrays, all host ports should be utilized. Nothing
requires you to use all four ports for a given server. Figure 1 below shows the high level
connection details with all four controller paths utilized. However, there are no technical
architectural or functional features in the MD32x0i/36x0i that requires you to use all four ports
for a given server.
At the hypervisor level, there is no ESX-driven requirement, for either performance-related or
pathing-related implementation that requires multiple IP-address subnets. More specifically, all
iSCSI ports can reside on the same subnet. The ESX native multipathing (NMP) driver is able to
use all paths on the same subnet, so if you choose a RR pathing policy, these ports will all be
used for Round Robin. Depending upon the workload needs, as shown in Figure 1 below, you
certainly could only use two of the ports if that meets your IO <throughtput> requirements.
Later in this whitepaper we include an in-depth explanation regarding why one might want to
use multiple subnets.