RightScale CEO Michael Crandell recently explained to a journalist at The Register that "the hybrid cloud is a little bit like sex in high school — everybody is talking about it, but not everybody is doing it." While that may be true for the IT industry as a whole, Michael also rightly points out what we in the RightScale Professional Services trenches see everyday: The majority of RightScale customers (87%) are leveraging the services of two or more cloud providers, which are often a public cloud and a separate, single-tenant private cloud.
Our customers run the gamut from the classic “two guys in a garage” success story all the way up to large enterprises, and many have already started to look at the private cloud for a lot of different and interesting use cases. What was once seen as an interesting science project has now been proven as a viable (and scalable) IT delivery model. Some companies are building “IT vending machines” that allow their internal users to quickly spin up pre-configured software stacks for use in siloed test and development environments.
Others are using their private cloud in combination with one or more of the public cloud offerings to create a hybrid cloud in which workloads are assigned to one side of the fence or the other depending on the characteristics of the applications they are looking to deploy.
And as the cloud outages over the last year or so have sent companies searching for viable disaster recovery solutions, the hybrid cloud has proven time and again to be up to this task for many of our customers.
In my role within the Professional Services group, I have the opportunity to talk to lots of our customers, and the trend continues to point more and more down the private/hybrid path. And as each of these customers have their own distinctive application workloads, they also bring to the private cloud their own unique set of requirements, desires, hopes and dreams. And while we in Professional Services do our best to say yes (as opposed to our Sales guys, who *always* say yes :) ) there are technical considerations that need to be considered, along with the normal practical ones like budget and schedule.
The companies I see that are the most successful at building private clouds are doing it for a few typical reasons, the first of which is workload/infrastructure interaction. In this case, the private cloud is being built for a very specific workload, and the hardware being procured is appropriate for the task(s) at hand. For example, one of our customers has an application with very high CPU demands. To most efficiently handle this workload in their private cloud, they are purchasing high-end compute hardware that has the horsepower to service the high demand. While there are public cloud resources out there that might be able to handle the load (such as cluster compute, high bandwidth instances), the quantity of the data involved is also quite substantial, and the cost (and latency) of pushing this data around, along with the security complications of shipping this data across the public Internet, has shown that the public cloud is a more costly solution over the long run.
Other customers have applications that demand (well, “very strongly desire”) high IOPS rates from their storage devices, and anyone who has run benchmark tests on the current selection of cloud-based storage systems knows that “your mileage may vary” doesn’t just apply to your choice of automobile and driving style. In order to get a higher (and much more consistent) IOPS rate, some customers are building out their private cloud with solid-state drives (SSD) for the nodes that require super-fast reads and writes. For the other parts of their application with more pedestrian needs in regard to storage, they have more traditional disk subsystems in place, and launch nodes in the appropriate pool of resources (segregating by zone or cluster).
Still other customers are using the private cloud to meet their security compliance needs. Since all the data stays within their own infrastructure, and never even sniffs the public Internet, they are able to run cloud-based workloads that still meet all of their stringent regulations. Basically, if you can keep your packets local and eliminate the need for them to travel out into the Wild, Wild West of the public Internet, you are going to have happier packets and unhappier bad guys who are out looking for those packets.
All of these use cases, the considerations surrounding them, and lots of other related and relevant bits can be found in our recently published white paper on architectural best practices for designing private and hybrid clouds.
Much like sex in high school, the prospect of having a private or hybrid cloud can be exciting, but if you’d never done it before, it can be reassuring to have that mentor who can tell you what to expect. We’re here when you need us. ;-)