Although the public cloud is often an integral part of an organization's disaster recovery initiatives, there are still plenty of organizations that rely on in-house resources for their disaster recovery efforts. There are both advantages and disadvantages to sticking with on-premises disaster recovery. For the purposes of this article, disaster recovery is defined as the ability to fail a workload over to an alternate location, whether that location is a separate cluster in the data center or the public cloud.
The most obvious disadvantage to on-premises disaster recovery when compared with the cloud is that the DR resources are vulnerable to data center-level disasters, such as fire or flood. If, on the other hand, an organization has the ability to fail workloads over to the cloud, then those workloads could remain online even if the company's data center is impacted by a catastrophic event.
Even so, there are a few advantages to keeping DR resources running on premises. One of the biggest advantages is predictable cost. Public cloud providers bill their subscribers for the compute, storage and network resources that they consume. The charges are usually based on complex pricing models that can make it difficult to predict how much it will ultimately cost to run a workload in the cloud. So, with that in mind, just imagine what would happen if all of an organization's on-premises workloads were to fail over to the cloud. The cost would be significant, but the actual cost would likely remain a mystery until the bill arrived.
Another potential advantage to using on-premises disaster recovery is the avoidance of a loss-heavy failover. In order for a workload to be able to fail over to the cloud, the workload's data must be replicated to cloud storage. The replication process is commonly asynchronous, with batches of data being replicated every few minutes. If a catastrophic event were to occur, several minutes' worth of data could potentially be lost, depending on the type of replication that is being used. Conversely, systems running on premises can usually replicate data in near-real time, greatly reducing the chances of data loss.
One more advantage to on-premises disaster recovery is that there is less latency involved. If a workload that had been running in-house suddenly fails over to the cloud, then client sessions will have to traverse a WAN link in order to use that workload. Depending on the speed of the link, users may find that their applications run slower than usual.
When it comes to cloud DR, there's still room for growth