storage snapshots — either on premises or in the cloud — can offer protection. In most cases, snapshots record the current state of the storage system, log any changes that are made into file descriptors and store any changed data incrementally.
In the event of a problem, cloud admins can roll back to a snapshot from a specific point in time and restore the storage system to that prior state. Any changes after that time will be lost, so provide some sort of journaling for critical files. Another alternative is to continuously snapshot data, which means that every write operation is snapshotted to enable rollback to near-real time.
Storage snapshot frequency
Admins can select from a variety of snapshot windows, from minutes to hours. The frequency they choose will depend on the workload.
For instance, because virtual desktops are not particularly volatile, organizations can snap them every hour or so. In this case, the loss of work wouldn’t be as detrimental, especially since commonly used apps in these environments, such as ERP or Exchange, have their own storage systems.
Databases, on the other hand, tend to have critical data and high volatility, which make them better candidates for continuous snapshots. When storage snapshots are continuous, they keep a permanent record of all changes and can be a useful forensic tool during an audit or when malware is detected. A good rule of thumb is that if you can’t afford to lose the last few entries, continuous snapshots make more sense.
The shortcomings of storage snapshots
Whether on premises or in the cloud, storage snapshots can introduce complexity. To build the desired recovery image of a file, an admin must tread through layers of changes, which is particularity time-consuming with databases and other volatile files. This overhead also applies to normal read operations and also adds time to writing changes out to storage.
In addition, there’s a lot of bandwidth involved in these extra I/Os. And while the speed of solid-state drives has allowed applications to support snapshots without taking a significant performance hit, organizations often have to pay a premium for SSD-based cloud instances and storage.
Lastly, snapshots do not provide data integrity guarantees. If the file system crashes or a natural disaster hits the data center, you will lose your data, so proper replication and geodistribution are essential. Also, since old data is kept as new data and takes up space on the volume, storage costs in snapshot file systems will be higher. Periodic cleanups might be necessary.