To many in IT, the elastic nature of cloud computing promises a world without resource restrictions. This perception is not only incorrect, but potentially dangerous in regard to the ability of enterprises to maintain critical business service levels for applications running within public cloud environments. Failing to (resource) plan will result in a plan that likely fails.

• There is an industry perception that cloud elasticity makes the need for capacity planning obsolete.

• Capacity "on demand" doesn't mean instantaneous capacity; it can take minutes, or much longer (depending on data placement), to initialize a cloud-based resource.

• Public clouds are likely to be designed like other shared public facilities — for example, airlines booking more seats than capacity, where oversubscription of services becomes a standard practice, with the potential to affect service levels adversely.

• The network's contribution toward application performance bottlenecks may increase due to its inelasticity, impacting applications unevenly.

• Workload placements will have to factor in cost and compliance issues.

• Elevate capacity planning from a tactical to a strategic process, due to its increasing business impact.

• Assign a capacity-planner role that includes cloud-computing environments as part of the person's responsibilities.

• Look for tools that go beyond simple resource allocation planning and modeling, and include cost, quality of service, compliance and other factors for cloud-computing environments.

• Use monitoring and testing tools, as well as interactions with development teams, to establish criteria regarding application architecture "fit" for cloud deployment.

Cloud computing enables enterprise IT organizations to expand the methods with which they can deliver critical IT services. As with any shared resource, there is no such thing as infinite scalability. The increasing popularity of cloud computing will likely result in increased resource contention as cloud providers rush to stay ahead of demand. In the meantime, we will likely see cloud providers act more like airlines, telcos and electric utilities, as they seek ways to shape and manage demand, requiring enterprises using these services to better understand concepts such as "yield management" (see Note 1), as well as providing enhanced planning for consumers of these services.

The public cloud-computing environment may become a victim of its own success, due to the increasing demands being placed on this new service delivery channel. We expect to see increased delays in starting up requested resources, thus impacting the IT organization's ability to provide "capacity on demand." End users may feel compelled to procure more-powerful compute instance models to maintain acceptable performance. Finally, restrictions may be placed on resource consumption.

Shared-service entities in other industries have had to develop strategies to meet increasing demand for their capabilities. Gartner believes that the cloud-computing market will face its own reckoning in this area in the not-too-distant future. Two scenarios are likely to emerge to meet increasing resource demands: (1) cloud service companies overprovision their resources to manage unpredictable demand (a costly, not to mention wasteful, exercise), or (2) a more likely scenario is service oversubscription by design (as is done in other industries, such as airlines that overbook seats), which may lead to periods of capacity constraints and increasing the likelihood of SLA breaches.

While compute (and storage) availability can usually be rapidly replenished (or reallocated) by cloud service providers in the face of demand variability, network bandwidth is largely fixed. Virtual LANs (VLANs), for example, help with isolation, but they do not necessarily enable the dynamic reassignment of bandwidth. And there will always be limits at the edge of these providers' infrastructures and into the Internet's "middle mile" in terms of available bandwidth. WAN acceleration devices, content delivery networks (CDNs) and other mechanisms may help, but they might ultimately provide only temporary relief. Increasing latency, and perhaps "jitter," could result in application behavior that is not easy to account for, let alone plan for.

As challenging as the resource availability issue is, there are other ramifications of moving to the cloud that need to be evaluated. Resource costs vary by provider. We've already seen the introduction of spot pricing (see http://aws.amazon.com/ec2/spot-instances/), and other market-oriented, demand-shaping mechanisms are expected. We may even see more-dynamic pricing, such as in the airline industry, in an effort to optimize cloud provider infrastructure utilization. Eventually, we may even see the rise of horizontal supply chains of cloud service providers to offer relief where specific resources become a bottleneck. Yet, "compute units" are not all the same; therefore, planners have to be able to analyze the differences — including for the associated SLAs. As a consequence, we envision the need for tighter synergy of processes and tools between capacity planning and IT financial management (ITFM) activities.

In addition to cost-related issues, there are nontraditional capacity (and resource) planning inputs, such as security and compliance, that also need to be taken into account. Cloud providers have widely varying support for industry-specific regulations (see http://aws.amazon.com/about-aws/whats-new/2009/11/11/aws-completes-sas70-type-ii-audit/). These regulatory compliance or security issues may impose additional resource and performance constraints on the enterprises, requiring their inclusion as part of the planning exercise with respect to workload placement.

Finally, an assessment needs to identify the ease associated with shifting workloads from one provider to another. There are not only cost ramifications to this (i.e., the bandwidth cost of moving the data), but also potential infrastructure compatibility and functionality issues.

Clearly, we are beyond the era where we planned to maximize the efficiency of the procurement of large "chunks" of resources, such as mainframes. Today's environment allows us to purchase resources online "by the drink," but the demand may be such that it is more difficult to determine the time it takes to fulfill the request. We still need to understand potential workload demands, but now we have to do this against an expanding array of providers. Being able to understand your business demand is important for another reason: No provider will notify you that your organization can make do with fewer resources. In a time where there are no "de jure" standards for cloud-computing units, we need to be able to compare and contrast offerings — some of whose pricing may become more dynamic. Interestingly, a few services are starting to emerge that might act as additional inputs to the capacity (resource) planning process. CA Technologies has partnered with Carnegie Mellon University to develop the service measurement index (SMI). SMI aggregates a broad array of key performance indicators (KPIs) for popular cloud services, and thus may become another input into future capacity planning products.

Also, let's not forget perhaps one of the most critical questions that a planning effort must answer: "Is the application or service (or just parts of it) a good candidate for running in the public cloud?" The answer requires a fundamental understanding of the application's architecture to assess its susceptibility to nondeterministic environments. In addition, analysis needs to be performed to determine whether or not an application will efficiently use all the resources for which the organization is paying. Because there are few tools available to help discern this, the capacity planner likely will have to interact more closely with testing and development organizations to establish corporate criteria in this area. Thus, a capacity-planner role becomes a key requirement for enterprises.

The good news with public (and even private) clouds is that IT organizations have more choices from which to source and/or run critical business services. This enhanced freedom of choice brings with it more responsibilities for the IT operations team, and specifically for capacity planners, in terms of ensuring effective end-to-end delivery. Capacity, or more specifically IT resource planning needs to move beyond its past tactical or clerical perception and take its rightful place as a critical element in a company's overall IT service management (ITSM) initiative.