Cloud Computing providers, including the private sector and fee-for service public sector, are providing Government-specific infrastructure and platforms and are beginning to differentiate their services by providing pricing tiers proportional to availability SLAs. This trend appears to exploit the concept that mission-critical applications to be migrated to the Cloud have not properly considered Disaster Recovery (DR) and Continuity of Operations (COOP), and should therefore be willing to pay considerable premiums to operate in a Cloud Computing environment with availability equal to or better than an Agency’s own data center.
The growth of high-premium high-availability Cloud Computing services indicates that we favor picking up applications from their legacy environments and executing a wholesale forklift to the Cloud without additional consideration for DR and COOP. While this is expeditious and appropriate for pilots and technology familiarization, it fails to exploit the nature of the Cloud as a commodity and perpetuates itself with continual reliance on higher-cost high-availability services.
Proponents of this approach will say that this is a temporary condition caused by vendor lock-in that will dissipate when data portability and cloud interoperability are achieved. In reality, any Agency that has database, ERP, or CRM is as locked-in to their software as they are to the point solutions that limit the brands of computing platforms that they purchase. We never speak about being able to instantly replace one COTS software system with another because we have data portability and software interoperability between the old and new, yet we are told that these are necessary elements to achieve ubiquitous Cloud adoption. While this is academically interesting, the simple truth is that Cloud vendors have almost no reason to support or implement cloud interoperability, and the reality is that it may never happen at all. Service Oriented Architecture is the appropriate architectural framework for achieving efficiency and agility via interoperability, and adopting Cloud alone will not help achieve these any more than a five 9s availability guarantee will keep your application running.
So when looking to deploy mission-critical applications to the Cloud, I would no more recommend deploying instances to any single provider than I would recommend deploying them to any single data center, whatever the availability guarantee. DR and COOP are the responsibility of the application owner and the appropriate considerations must be made during all phases of the lifecycle. Application migration to the Cloud is an opportunity to reexamine how availability is implemented, executed, and tested; it is not a convenience to transfer responsibility for availability to an infrastructure or platform Cloud vendor.