Information Technology Reference
In-Depth Information
set thresholds and actions, and then extend LUNs and the VMFS datastores using the new
vSphere VMFS extension capability, or grow NFS datastores.
When thinking about availability
◆
Spend the bulk of your storage planning and coni guration time to ensure that your design
has high availability. Check that array coni guration, storage fabric (whether Fibre Channel
or Ethernet), and NMP/MPP multipathing coni guration (or NIC teaming/link aggrega-
tion and routing for NFS) are properly coni gured. Spend the effort to stay up to date with
the interoperability matrices of your vendors and the i rmware update processes.
Remember, you can deal with performance and capacity issues as they come up nondis-
ruptively (VMFS expansion/extends, array tools to add performance, and Storage vMo-
tion). Something that affects the overall storage availability will be an emergency.
◆
When deciding on a VM datastore placement philosophy, there are two common models: the
predictive scheme and the adaptive scheme.
Predictive scheme
Create several datastores (VMFS or NFS) with different storage characteristics, and label
each datastore according to its characteristics.
◆
Locate each application in the appropriate RAID for its requirements by measuring the
requirements in advance.
◆
Run the applications, and see whether VM performance is acceptable (or monitor the HBA
queues as they approach the queue-full threshold).
◆
Use RDMs sparingly as needed.
◆
Adaptive scheme
Create a standardized datastore building-block model (VMFS or NFS).
◆
◆
Place virtual disks on the datastore. Remember, regardless of what you hear, there's no
practical datastore maximum number. The question is the performance scaling of the
datastore.
Run the applications and see whether disk performance is acceptable (on a VMFS data-
store, monitor the HBA queues as they approach the queue-full threshold).
◆
If performance is acceptable, you can place additional virtual disks on the datastore. If it is
not, create a new datastore and use Storage vMotion to distribute the workload.
◆
Use RDMs sparingly.
◆
Our preference is a hybrid. Specii cally, you can use the adaptive scheme coupled with (start-
ing with) two wildly divergent datastore performance proi les (the idea from the predictive
scheme), one for utility VMs and one for priority VMs.
Always
read, follow, and leverage the key documentation:
◆
VMware's Fibre Channel and iSCSI SAN coni guration guides
VMware's HCL
◆
