This article originally appeared in the March 2011 issue of INTERNET TELEPHONY.

I would consider the topic of cloud bursting to be one of the more advanced cloud computing concepts, but also one that shows the most cloud functionality and return. As a relatively new idea – although almost all cloud ideas are still new concepts – cloud bursting certainly is getting a lot of conceptual architecture time. It’s something that most people can grasp as an immediate benefit of cloud computing.

In a nutshell, cloud bursting is the idea of using cloud-based resources as an extension of existing resources. The principles of cloud computing – limitless scale, dynamic provisioning, on-demand resources, etc. –still are present with cloud bursting, but they’re only invoked after resources within the physical data center are exhausted. For example, an IT department may host a website in its own data center, but can only handle approximately 10,000 connections per minute. During a spike that drives connections over that threshold, IT can send the additional connections to an on-demand cloud environment dynamically and seamlessly. When demand dies down, the cloud environment is de-provisioned and the on-premises web hosting manages the normal load.

Ideally, cloud bursting provides the best of both worlds: It allows the enterprise to maintain ownership and control over its application on-premises while having a fail-safe off-premises solution when needed. But there’s a trade off: That flexibility comes with a more complex computing environment to manage. First, there’s building the cloud bursting environment and keeping it current. Since the bursting environment is only used based on need, it’s conceivable that it may not be used for months at a time – great for the budget but a challenge to keep current. Every change to the on-premises application needs to be replicated to the mostly dormant bursting environment. While many IT departments are struggling with virtualization challenges such as managing virtual machines and virtual sprawl, adding an extra off-premises virtual environment can compound management challenges.

Managing the bursting environment as an extension of the on-premises data center is one of the most important – and difficult – design factors. Much of the pre-planning comes down to choosing a provider that understands the bursting environment is basically a disaster avoidance solution rather than a full managed cloud. This minor difference is important: A cloud bursting environment is basically always on deck, but may never see any action. It’s the special team of the data center world. The cloud provider needs to offer an environment where that team can be called up at any time in an automated, programmatic way, from the management platform as if the resources were on-premises. It does no good to keep resources at the ready if you can’t spin them up on your terms. Choosing a provider that can tie into your event management system is a make or break decision.

That leads us to the third consideration for cloud bursting: integration. By definition, cloud bursting is tied to another, existing data center environment. Applications only need to burst into the cloud when there’s a demand that is exhausting existing resources. The bursting event depends on a threshold trigger that starts the event, a workflow that can flip the switch to turn up the burst environment, and a network that can manage connections as they move from on-premises to off-premises. All of that logic needs to come and be based on the existing infrastructure. Bursting is triggered based on a dynamic change in demand, but that demand will come from the existing infrastructure. It’s critical that both on- and off-premises environments be integrated so they can function as one network delivering applications to users.

These are just three major considerations for cloud bursting; there are many other smaller but equally critical components to a successful bursting architecture. Virtual platform choice, VM management, dynamic routing, application monitoring, SLAs, centralized message bus, dynamic DNS – the list goes on and on. There are some cloud providers that are well ahead of the curve and are offering pre-packaged cloud bursting solutions. These can be excellent choices provided they match your current management environment and can integrate with your data center out of the box, or the provider offers a la carte solutions that can be compiled to match your specific environment. Other providers are offering internal bursting options for traditional managed and hosted customers, building an architecture that allows those customer environments to burst as needed but keeping everything localized to the provider’s network – also excellent options if you’re already using an MSP or hosting provider that offers those solutions.

Cloud bursting may well be what propels cloud computing into IT’s standard toolbox, but critical yet simple decisions during the design process with your cloud provider will make all the difference in the world between a seamless disaster avoidance solution vs. one that does everything but avoids a disaster. 

Alan Murphy is technical marketing manager of management and virtualization solutions with F5 Networks.