Proof Point: Example Clustered Microsoft SQL Server Configuration on the HP ProLiant DL980

5
The main public LAN connection for each cluster node consists of two NIC ports teamed together, one from
port 2 of an NC360T in slot 1, and the other from port 2 of an NC360T in slot 7. Slot 1 is a PCI-e 1.0 slot in
the DL980, but still has ample bandwidth to keep up with a dual gigabit ethernet adapter. Port 1 on each NIC
card is not used for the purposes of testing. Two distinct cards are used, rather than two ports on one card,
to provide redundancy against the theoretical possibility of one slot failing. During testing, one or the other
LAN cable (but not both at once) of the main public LAN is periodically disconnected and reconnected, to
verify that the NIC teaming tolerates the fault and recovers properly.
The HP Network Configuration Utility is used for the teaming. Default settings are used, except that the
Team Type Selection is Switch-assisted Load Balancing with Fault Tolerance (SLB).
Embedded NC375i port 1 on each node is used for the cluster heartbeat. This connection is not teamed, as
failure of the heartbeat would not cause catastrophic failure of the cluster. The heartbeat is on its own
subnet, distinct from the subnet of the main public LAN. However, both subnets are physically connected to
the same network switch, an HP ProCurve 2900-48G. There is only one network switch, as network switch
redundancy is not part of this test case. Embedded NIC ports 2, 3, and 4 are not used.
Storage Layout
The operating system boots from two 146 GB disks in the embedded drive bays of the DL980. These disks
form a RAID-1 volume through the internal p410i Smart Array. Note that the Microsoft SQL Server 2012
executable also resides on this volume.
All cluster shared storage volumes are on HP p2000 storage enclosures, connected through fibre channel.
Figure 3 Storage Layout - User Database Raid 5
The first P2000 enclosure contains two arrays. The array that holds the User Database volume is RAID-5
with a hot spare disk, so that it can recover automatically from a disk failure. In fact, this functionality is
verified during testing. The second array holds backups of the user database.
Each time the test is run, the database is restored from backup, to insure a consistent known state. As it
takes much longer to recreate the database from scratch, the backup volume is valuable, and protected with
RAID-5. Each array also contains an additional small volume for test purposes. These small volumes are
slightly different in size, for easy identification.
Figure 4 Storage Layout - TempDB Raid 10