Part II of a Two-Part Series
In part I, I covered the basics on why Snapshot technology is the most frequently employed technology to backup data at a specific point-in-time. In this second submission, I’d like to focus on how Overland has incorporated snapshots into its ULTAMUS® RAID to support data protection best practices.
Snapshot functionality is a very integral component of Overland’s ULTAMUS RAID firmware, which runs in the RAID controller rather than on a host—maximizing performance while ensuring full availability to data as applications continue to process.
As shown in the diagram below, ULTAMUS RAID snapshots employ a Copy-On-Write (COW) process, meaning the snapshot copy and primary data share the same data source, except when new data is created. The snapshot keeps the unchanged file and primary data keeps the updated file.
A COW snapshot produces a result similar to full volume copy, but a snapshot is captured in far less time—and requires far less disk space—than required for a full volume copy. Once a snapshot is taken, a server writing to the source logical volume transparently causes pre-defined segments of the source logical volume to be copied to the snapshot before allowing the write to continue, thus preserving the original data on the snapshot.
When in operation, servers continue to read from the source volume and write to a special reserved area with the user unaware that the snapshot has occurred. Once the snapshot is released and the source volume is "unfrozen," data that was written to the source volume while the snapshot was active is re-integrated into the source volume. This activity is completely transparent to the user making snapshot clean and non-intrusive, which, according to my research, is required in most cases.
Clear as mud? If you are like me, visual representation works best…
It is important to note that a COW snapshot is not what the acronym implies. In fact, it is the exact opposite as a COW snapshot is not a complete physical replication of the original disk requiring the same amount of space as the original volume. A COW snapshot is only a virtual copy, as all data written to the source volume will still reside on the source disk.
This implementation requires a minimal amount of storage capacity, typically between 10-to-20 percent of the source logical volume (dependent on data change rates, of course), enabling it to quickly generate and efficiently store several snapshots within roughly the same space required for a single mirror or copy of the volume.
Real-world access and use of snapshot varies. For example, using products from Overland, a backup server can mount the snapshot volume from ULTAMUS® RAID and backup the frozen data to a REO®, NEOTM or ARCvaultTM while the application server continues in production. This same snapshot volume can also be used to roll-back the volume to a previous state or be used for a test environment when validating a process within a Business Continuity (BC) plan.
With all that said, snapshots, undeniably, are vital in any BC plan as they provide the most efficient, flexible (as stated above by the numerous key benefits) and reliable method of data protection and restoration by providing administrators with a copy representing an entire volume at a given point in time.
--
Kevin Wise
Product Manager