One of the biggest criticisms of using primary storage for DR is cost. Primary storage can be expensive, and using it in your DR plan essentially doubles your storage requirement. Therefore, it is essential that DR Ready storage use as little space as possible and provide granular protection policies. Let’s take a look at the ways that it can do that.
One of the primary ways to reduce actual storage use is to use thin provisioning. This feature allows customers to provision much larger volumes than they actually need, while the storage product only uses blocks that they actively use. For example, a user can create a 10 TB “logical” volume that only contains 1 TB of “actual” data. This feature saves space, aka cost and makes management easier, as storage administrators do not have to continually re-provision their volumes as they need more space.
The next feature that is important, both for data protection and for space efficient use of storage, is snapshots. Snapshots are an essential part of storage and often combined with replication for disaster recovery. Replicated storage without snapshots can be just as useless in a disaster as no storage, because the replication would also replicate any kind of logical corruption such as a black hat attempting to damage your data. Snapshots allow you to go back in time prior to any malware or other corruption to your system.
A snapshot – as the name implies – is a view of your storage rather than an actual copy. Subsequent snapshots take up very little space, as they only save the changed blocks and pointers since the original snapshot was taken. This is very good from a storage efficiency perspective, as one can have hundreds of snapshots without requiring a significant amount of additional storage to store them. In fact, storing 100 snapshots takes no more storage than storing a few data changes, because each snapshot only stores the blocks that are unique to that snapshot.
The one thing that one must keep in mind about snapshots is each snapshot is a virtual copy. Snapshots must be replicated to another destination in order to be able to use them in a disaster. A snapshot of your storage after it catches fire is as useful as a snapshot of your house after it catches fire. You must replicate snapshots for them to be useful in DR.
The DR Ready system should provide administrators with granular control over which of these snapshots gets replicated because of the additional efficiency and cost savings gained when done at the right abstraction level, as my colleague George Crump outlines in his entry “DR Ready Storage should be a Scalpel or a Saw?”
Two other technologies are compression and deduplication. They are mainstream now but confusion and hype still remains. It’s important to realize that the absolute space savings factor is not the full picture. Their savings differ when applied to different workloads.
Compression looks for redundant patterns within a given file or volume, and replaces those patterns with pointers. The easiest way to understand compression is to consider a large document that happens to have the same word repeated many times. A compression algorithm would notice the repeated word and replace all of the occurrences of that word with a single reference, saving a significant amount of space. If, however, that same word was repeated in another word processing document, compression would not notice that.
Deduplication, on the other hand, looks at patterns between volumes or files. In the example above, deduplication would notice that the same word is present in multiple files, and would replace all incidents of that word with pointers. The most applicable use of deduplication in primary storage is when storing VM images. Each VM image contains a copy of the operating system and one to many applications. These redundant parts of the VM can be identified and replaced with pointers, resulting in significant space savings for those blocks.
Each of these technologies have their role, and save a significant amount of space; however, mileage will vary depending on the workloads. DR Ready storage needs to counteract the cost of having a second storage system that is used only during a disaster. When all of these data reduction technologies are implemented, it is possible to have DR Ready storage with similar cost structure as maintaining an entire backup and recovery system. And there is no question that DR Ready storage is a much nicer thing to have than a traditional backup system.
Sponsored by Tintri