Data replication creates a copy of data from a production environment to a separate location and allows users to fail over to the secondary location in the event of a failure. It also ensures that the production environment itself can be restored in case of a problem.
In this Q&A, Greg Schulz, senior advisory analyst and founder of StorageIO, discusses snapshot, replication and other fundamentals and how organizations can benefit from their implementation.
Greg Schulz: Both backup and replication are part of data protection [that makes sure] there is a copy elsewhere. And when replication is combined with snapshot or some other form of creating a recovery point or consistency point, the lines between it and backup can be blurred.
The key is that replication by itself -- without snapshots or CDP [continuous data protection] or some form of versioning or consistency point -- is just a stream or mirror of the source. [By comparison,] a backup, snapshot or some other form of creating a recovery point enables going back to that point in time. Many backup solutions today are integrating with snapshots to manage those as data protection points in time similar to how they have done so for traditional disk, tape and cloud copies.
Are people typically using software for replication or relying on replication that comes baked into their storage arrays?
Schulz: If you talk to storage-focused or -centric people, the answer is usually storage system, appliance, storage virtualization or related software-based approach for applications or databases. [You] might be surprised how other forms of replication including copy/shipping log files are being done. In some environments, the replication or copy is all done via the storage system or appliance; in others, it's via some host-based storage -- including hypervisor-, operating system-, file system-, or database-based; while others use a hybrid approach.
What are the benefits of relying on array-based replication?
Schulz: Tighter integration with the storage system [is one benefit]. Hopefully, the storage system replication also has application plug-in agents for data consistency as well as complete data protection.
Schulz: [An organization can] avoid having to pay for the storage replication or complexity in managing it or having to rely on a specific vendor, not to mention the ability to use different vendors' storage.
What about using replication native to your virtualization platform?
Schulz: Yet another approach is to use replication and snapshots along with operating system and application integration on hypervisors with their guests. Some also leverage storage system offload to implement a hybrid approach.
Which one is best … depends on what you are trying to do or the limits and constraints of your preferred technology or tool. There have been and will continue to be some entertaining debates on the best place to do replication. While fun to listen to, the key is determining which is relevant to your specific needs. Storage systems can offload [workloads from] the servers or applications. On the flip side, other approaches can offload workloads from the storage or move away from a propriety approach.
And if you have extra CPU server cycles that you were going to consolidate anyway, you can avoid having to be storage-centric. Then there are the hybrids of doing [it] in an appliance, storage virtualization solution or via host-based third-party software.
How do you see replication being used today? Are organizations using it with snapshotting for a CDP-type approach?
Schulz: Replication should be used with snapshots, and snapshots should be used with replication. Snapshot and replication are better together and when used effectively together can be a great approach for implementing data protection, including as an alternative to traditional backup/restore.