One of the new features in the recently announced VMware vSphere 5 is Profile-Driven Storage. Duncan Epping has a great blog post that explains how the new feature works. Combined with the vSphere Storage APIs for Storage Awareness (VASA), this feature makes it easier to find an appropriate datastore when provisioning or moving a VM.
VASA allows a storage array to export information about its configuration, such as RAID level, replication, or the presence of solid-state drive (SSD) storage. The storage administrator can then create profiles based on these capabilities. For example, “Gold” datastores may have RAID 10 with replication, while “Silver” datastores are unreplicated RAID-6 LUNs. When it’s time to provision a VM, the user interface allows the administrator to assign a storage profile and limit the available choices to datastores matching that profile. This decision can also be driven by nontechnical considerations, such as organizational ownership; for example, some datastores could be tagged for test and development use.
Profile-Driven Storage is just one of the storage enhancements available in vSphere 5. Storage is clearly a pain-point for VM deployments, and VMware is actively trying to find solutions that will help its customers. Classifying and providing more information about the available storage is a good thing, but it’s not a complete solution. VASA and the UI enhancements are tools that help manage the complexity of legacy storage, but do not eliminate it.
The storage requirements of a VM should be properties of the VM itself, not of the datastore it resides on. Virtual machine administrators must be able to define criteria like data protection levels, application performance requirements, or performance isolation (for multitenancy) on a VM basis. At that point, the storage array should implement the desired behavior, independent of what datastore holds the VM’s data. If the administrator doesn’t initially understand the VM’s requirements (as is often the case), the storage should automatically adapt, rather than require an explicit transfer to a different tier.
The Tintri VMstore provides a single datastore with both top-tier Flash-based performance for VMs that require it, and inexpensive SATA storage for VMs that don’t. As we roll out new features like replication, our VM-aware file system will allow us to target individual VMs within that datastore rather than requiring a data-placement decision to also become a capability decision. Our approach eliminates the need to set up and configure multiple tiers of storage, and instead provides a single flexible datastore that automatically adapts to application requirements.
In the long run, we believe this is the only way to manage the complexity of VM storage. Technologies like Storage DRS are a step in the right direction—from individual storage arrays to pools of storage that are automatically managed. But matching storage resources to VM behavior can best be done by a VM-aware storage system, which understands and adapts to the requirements of virtualized applications.
Unique control with VM-level actions for infrastructure functions including snapshots, replication and QoS make protection and performance certain in production, and accelerate test and development cycles.