White Papers

Deploying Microsoft SQL Server 2008 based Decision Support Systems using Dell EqualLogic 10GbE iSCSI Storage 9
Since DSS environments are highly read intensive and require large storage capacities, it is a best
practice to use RAID50 with EqualLogic PS Series storage arrays. RAID 10 is also recommended for
environments that include a mix of both read and write operations. For more information on setting
RAID policies see Table 1: Matching RAID levels to DSS I/O componentsTable 1, in Section 4.5.2.
4.1.3 StoragePoolConfiguration
In each test configuration we included all member arrays in a single EqualLogic group. The storage
arrays under test were managed using a single instance of the Group Manager user interface.
We created two storage pools in each test configuration. One pool was used for hosting the SQL
Server® data volumes, including the tempdb volumes. We used a second pool for hosting the SQL
Server® transaction logs. The transaction logs were physically separated from the DSS database disk
drives for high levels of data protection and recoverability.
DSS workloads are primarily read intensive and they typically generate minimal transaction log activity.
Thus, in smaller DSS database environments using Dell EqualLogic storage arrays the transaction log
volumes can be hosted on the same pool and disk drives as the data volumes without sacrificing
performance. However, if high levels of data protection or recoverability are required SQL Server
mirroring or replication solutions can be considered.
For DSS solutions that require high levels of data protection and recoverability, it is a best practice to
host data and log volumes on separate disk drives and pools, or consider application data mirroring
solutions such as SQL Server mirroring or replication.
4.2 SAN Design
You should consider the following best practices when configuring the SAN for a DSS solution.
4.2.1 UseRedundantSwitchPathsBetweenServersandStorageArrays
In our test configurations we used redundant switch paths in the SAN to ensure uninterrupted access
to storage controllers in the event of a switch or interconnect cable failure or a scheduled switch
downtime for maintenance. Figure 4shows how we used two switches to create redundant
connection paths between the server NICs and storage array controllers. This example illustrates the
following best practices to follow while designing an iSCSI SAN for a DSS solution:
 Equally distribute the NIC ports installed on the server for iSCSI SAN connectivity across the
redundant switches.
 Equally distribute the Ethernet ports from each controller of the storage array across the
redundant switches.
 Interconnect the two switches with sufficient number of inter-switch links (ISLs) to support the
workload under test.
In our testing we found that a link aggregation group (LAG) consisting of four 10Gbps ISLs between
the Dell PowerConnect 8024F switches was required for the workload under test. It is a best practice
to provision the LAG bandwidth based on the throughput requirement between the servers and arrays
plus I/O workload characteristics. The required number of switches in the SAN design depends on the
number of servers and the number of storage arrays required to support the workload.