Is it possible that the public cloud can give data centers what they’ve always needed? Is it possible to have very short RTOs and short RPOs by using someone else’s infrastructure to do the job? This is looking more and more likely to be the case.
A successful disaster recovery is one that doesn’t experience significant downtime, doesn’t experience significant data loss, and continues to serve the appropriate applications at roughly the same performance. Accomplishing such a recovery requires a number of activities and conditions.
Obviously, one must have an up-to-date backup of the data center needing replacement. The traditional method of accomplishing a backup is a nightly backup, which unfortunately takes too long to create and is not done often enough. This results in a significantly longer recovery time and a significant loss of data during outage. Yes, it is true that you can use database transaction logs to restore some of the data created since a backup was taken – if those transaction logs are still available. In a true disaster they are no longer available.
This means you can accomplish an up-to-date backup only through continuous backups to an off-site location. This is why, prior to the advent of the cloud, most people serious about disaster recovery used replication to another site for mission-critical applications. The problems with the replication method are legion, but include the cost of a second data protection system and the day-to-day cost of operating and maintaining another data center.
Aside from the monumental cost of maintaining a data center whose purpose is only to take over in a disaster, there are technical challenges with this idea. If the data center is to take over for production, the data center needs to mirror production. The servers in the hot site, as it’s called, need to be the same kind of servers with roughly the same compute power as what we find in the production data center. For cost reasons, this was almost never the case in a typical hot site. At best, the hot site was given “hand-me-downs” – the previous versions of servers used in the data center. Physically migrating servers from the production site to the hot site was a costly process and it came with the understanding that the hot site would never be able to truly take over for the production site in a full disaster. There simply wouldn’t be enough processing power.
Cloud computing companies and disaster recovery-as-a-service products solve all of these challenges. A good DRaaS product can provide both operational recovery and disaster recovery with a single off-site backup, which can be continuously updated throughout the day. The copy in the cloud can be immediately available (offering a very low recovery time) and can be kept in sync with the production data center (offering a very tight recovery point).
But the truly beautiful thing about cloud-based disaster recovery is that the cloud data center can offer the recovered company exactly the same (or more) computing power it had in its production data center – without the cost of having to actually maintain that hardware. Companies using a DRaaS service pay only for the storage of their data in the cloud, unless there is a declared disaster. Then, their bill also includes compute. This allows them to maintain a dream data center, continuously updating the VM definitions to reflect what they need in a recovery, without incurring any of the costs.
We discuss this and other aspects of cloud-based recovery in our on demand webinar “DRaaS Best Practices – 5 Critical Recovery Steps”.