Is virtual machine performance better when storage is on the local server or a storage area network (SAN)?
Workloads, like virtual machines, are typically unaffected by storage location, as long as the storage location provides adequate capacity and performance.
Local data storage (disks on the servers themselves) may yield somewhat better performance simply because a disk interface like 6 Gbps serial-attached SCSI (SAS) is dedicated to fast storage tasks. Also, local storage doesn't rely on sending data over the network and therefore is immune from network bottlenecks and disruptions.
The problem with local data storage on virtualized machines is that migrating a workload between servers would also require migrating the workload's local disk contents, which can be a time-consuming challenge. Centralized storage overcomes this problem by providing uninterrupted access to storage regardless of which server the virtual machine is on. The VM's storage location never changes even though the VM can begin running on a different server on-demand.
In addition, local storage is rarely used in enterprise environments because it is cumbersome to monitor and manage. Centralizing storage in a SAN or NAS eases storage management challenges and can even simplify backup and disaster recovery practices with replication features already within the storage subsystem. These are all important considerations for virtualized environments.
So, while local data storage may offer some performance improvements, centralized storage pools offer more features and are easier to manage. Which approach is best for your organization will depend on the size of your environment and how you use your servers.
This was first published in October 2013