Thursday, April 1, 2010

Storage Virtualization

As we are implementing Microsoft virtualization, more and more storage space are being used up rapidly. Another issue is storage availability. As we cluster up more VM hosts, major single points of failure still remain on the shared cluster storage. Even if storage can be fully replicated within a single site, any site-wide disaster (like flood, fire etc) can wipe off any data shortly. This is where storage virtualization comes into the picture. Wiki defines storage virtualization as the abstraction (separation) of logical storage from physical storage.

Let's take a look at the jargon used and how they can help solve the above issues, mainly on over-provisioning (that lead to high costs) & availability/DR related issues.
  • RAID: Some said RAID is the earliest form of storage virtualization, as a logical volume can span across multiple disks to prevent single disk failure.
  • I/O Multipathing (MPIO): In the event that one or more of these components fails, causing the path to fail, multipathing logic uses an alternate path for I/O so that the servers & applications can still access their data.
  • Remote synchronization: To eliminate storage as single point of failure, data across two separately located storage are replicated over the network on a per volume basis. It presents a single logical volume to the servers, although it may span across different storage boxes. It is also essential to implement multi-site failover clustering for Windows 2008 servers.
  • Thin provisioning: It is easier and less troublesome to extend a volume rather than shrinking it. Hence, most administrators tend to over-provision storage space for applications. To reduce wastage, thin provisioning allows administrators to provision a large volume but only a small fraction is actually allocated until the applications occupy more space over time.
  • Thin replication: You replicate a "thinly" provisioned volume. Only delta changes will be replicated across and save network bandwidth.
  • Point in time Snapshot: To simplify data restoration & DR recovery, periodic snapshots on the storage are taken over time. It allows you to rollback data to certain points in time.
  • Deduplication: When you implement server virtualization or VDI, most of the bits and bytes of the VHDs are identical. Deduplication further optimizes storage space by removing duplicated bits & bytes.

No comments:

Post a Comment