A recent study by 451 Research found that 20 percent of cloud users have moved at least one or more of their workloads from the public cloud to a private cloud. Now popularly known in our industry as “cloud repatriation,” the report also found that an additional 10 percent of public cloud users were planning to move some workloads from public to private cloud. In total, companies that were embarking on cloud repatriation planned to do so with 40 percent of their current public cloud workloads.
So, when does it make good business sense to move an application out of the public cloud and return it to an on-premises data center?
Typically, a move of an application or dataset is warranted when critical operational benchmarks are not being satisfied. Causes for this include but are not limited to inconsistent application performance, high network latency due to congestion and/or consumer-application distance, security concerns, working dataset size, or concerns about cost. Applications that are latency sensitive, have long-running I/O intensive periods, or have datasets which are large and require transport between various locations for processing are generally prime candidates for repatriation.
The benefit of cloud repatriation is the balance that is achieved within a large complex system of applications and datasets. As operational conditions change, so too must the balance of applications and data with regard to locality.
Now more than ever, we live in a software-driven world. That is, software provides the means by which we are able to move applications and data from one location to another with little to no downtime. Much of the advancement that is occurring in hybrid cloud software stacks is focused precisely on the issue of allowing workloads to move freely, either automatically or on-demand, across various cloud platforms and locations with minimal disruption.
Are there potential drawbacks to cloud repatriation? Of course. As with any IT decision, a poorly researched, planned and tested process is almost certain to cause pain for the end-user community or the company as a whole.
This is again where software comes into play. There are many software packages available today that can greatly assist with balancing the application-locality equation. In fact, many of the more mature cloud stacks are beginning to include these capabilities natively to provide an “easier button” for this type of operation. It’s important to realize that this is an operation which is not a once-and-done task, but a continually ongoing one which IT administrators are going to be faced with much more in the near term.
As we look toward the future, expect to see further abstraction of the public and private clouds, essentially reducing the “cloud” to simply a computing platform selected primarily based on location.