The forecast for cloud computing growth isn't the least bit murky. Demand projections are nothing short of stunning, with Morgan Stanley estimating that market spend on Infrastructure as a Service solutions will top $152 billion by 2022.
For an industry segment that was nonexistent at the middle of the last decade, this is amazing growth. Even the most cautious enterprises that were watching the cloud hype from the sidelines just a few years ago are now moving from their initial on-demand experimentation with such applications as test and development to leverage the cloud for more operationally important activities.
Scaling up and scaling out is not as simple as dynamically provisioning resources on-demand.
The tremendous revenue potential is driving virtually every hosting company and communications and IT solutions provider to invest in building out a cloud portfolio. The catch for providers is how and when to scale cloud infrastructure to meet increasing demand without overbuilding. Part art, part science, cloud capacity planning presents challenges to providers that need to figure in their own and their customers' needs.
Fundamental to the cloud's appeal is the flexibility and efficiency associated with the virtualized environment that underpins it. One key advantage virtualized environments have over traditional architectures is the cost benefit associated with consolidating computing power and storage capacity on fewer physical devices to lower both capital expenses and support costs. Cloud providers can then pass the overall cost benefits to enterprise clients.
Cloud orchestration and automation play crucial roles in this virtualized delivery model, supporting the kind of dynamic provisioning that makes it possible for clients to tap into just the compute and storage resources they need, when they need them. This further extends the cost efficiency of the on-demand delivery model the cloud facilitates. The challenge remains that the virtualized elements are still attached to fixed physical resources, such as CPU, memory and network bandwidth. Thus, scaling up and scaling out is not as simple as dynamically provisioning resources on-demand.
Instead, a cloud provider's success is tied directly to its ability to properly estimate resource requirements and to successfully scale its infrastructure without overcommitting and overbuilding. Because cost is such an important element in the competitive cloud equation, as are concerns about reliability and performance, providers need to have a firm understanding of their customers' expectations for Quality of Service as they architect their cloud platforms.
Projecting in a good way, despite demand-side unknowns
Unfortunately, one of the big challenges that remain for cloud providers, in terms of planning, is how to accurately project both short- and long-term capacity requirements at a time when, in spite of tremendous interest in cloud, most enterprises are still grappling with cloud strategy and have yet to fully commit to the model from a budget perspective. So how can a provider accurately estimate future capacity requirements when so many unknowns about the demand side remain?
The prepared cloud provider needs an accurate view of its current infrastructure and services portfolio, as well as the needs of its existing customer base. A cloud provider should have a clear picture of its existing capacity and know how much of that is being consumed now. A provider can also deduce a lot about future needs by understanding the immediate demands of its installed base and the buying patterns of that client list, both in terms of cloud computing services and more traditional hosting and colocation services. Providers should ask customers about their expansion plans and whether clients' resource requirements vary on a seasonal or cyclical basis.
More cloud capacity planning resources
Capacity planning is about Capex and SLAs
Amazon Marketplace addresses flexible cloud capacity planning
Cloud storage management: More than capacity planning
Cloud providers should also factor in their own expansion strategies when planning future capacity needs. At a high level, a provider should look at where its best new opportunities are in terms of new markets and geographies, and whether this expansion will require establishing infrastructure and operating resources in a new region. Providers considering selling into new vertical industries should also consider variable capacity requirements and other buying trends specific to that segment, as well as short- and long-term expectations for overall cloud investment.
Academics and developers are working on tools and techniques to help cloud providers use current information on capacity and usage, in conjunction with forecasting data, to formulate a model that estimates resource demand. This model would accurately scale providers' infrastructure to meet future customer requirements. Cloud capacity planning formulas should be able to break down demand on a component basis to project what discrete physical resources and software will be required going forward.
Cloud providers need to carefully evaluate their return on investment expectations for the future before they commit to any major capacity build-out. Vendors recognize the stresses and pressure cloud providers are under as they try to assess future infrastructure requirements. Many offer providers options that include revenue-sharing models and more flexible buying terms to lessen the up-front capital expense.
No matter what method cloud providers choose for capacity planning, they need to have some assurances about accuracy. They should find out whether specific tools and formulas have been used by other providers and what margin of error should be built into their estimates. Accuracy is absolutely essential, not just for the cloud capacity planning exercise, but for the efficiency of the provider's overall go-to-market strategy.