I am astonished by how little practical or empirical data exists on the topic of cloud bursting. A quick Google search on "cloud burst" or "cloud bursting" yields, well, not much - that is, short of "Men Who Stare at Goats" references, questionable YouTube clips and a campy '80s "art rock" video. To further mystify the topic, all of these data points really revolve around the dubious and Fringe-esque claims of "cloud busting" (notice the missing "R") - or making rain by tampering with clouds.
Yet, the concept of cloud bursting (with the "R"), or horizontal application scaling into the cloud (i.e. moving compute workloads into an on-demand resource pool to access additional capacity), has come up in just about every one of my conversations with enterprise clients. Why? Because this could be one of the fastest and most impactful ways for customers to harness the power of cloud computing to grow applications and respond to seasonal, cyclical, or ramping demands ... and it's really pretty straightforward if you have selected the right cloud provider.
I guess I shouldn't be too surprised about limited information on the topic, considering that cloud capabilities vary greatly from provider to provider (see our evangelism efforts on Not All Clouds are Created Equal). Therefore, there is not one easy three-step guide on how to cloud burst. But, hypothetically, if one were to exist it might look something like this - assuming you have your workloads already virtualized:
1. Define what you will be bursting. Of course, this relates to applications - but the important question is which layer within the application architecture is ideally suited to a cloud bursting use case? Is it the web, application or database layer within a traditional three-tier relational database-driven application stack, or are we talking about a flat file, NoSQL or big data bursting scenario?
2. Select your target cloud. This directive is tightly correlated to how you responded to the first step, since cloud service providers each handle tiered models and distributed data models differently. Most enterprises tend to prefer highly secure and high-performance cloud that makes it easy bring in workloads.
3. Convert your source images and upload. Based on the provider you have selected, it's time to bring in your image. Is it a VMDK, OVF, XenVM or something else? It's common that VMs in even Open Virtualization Format should adhere to some service provider-specific configurations. Tools like VMware Studio, Platespin and others can be used to convert workloads.
Now that you have identified your applications, chosen a provider and converted your image to interoperate with your cloud, as well as uploaded it, you are practically there! However, there are still several factors to consider, and cloud vendors handle these very differently:
- How will you launch your workloads? From a template, from a clone or from a dormant VM/instance?
- How will you connect, and how much data do you intend to push over this network connection? Is it a point-to-point network, MPLS, EVPL or VPN, and is it production data, metadata, sensitive data or management traffic?
- How automated should this solution be? An API can provide full automation, but will require coding and additional business logic in your applications. Is the cloud portal you have chosen easy enough to operate to take advantage of cloud bursting?
- How will the cloud handle your security policies? Does the cloud you have chosen have the governance and maturity you would expect for you data? Can you even bring your own policies into the cloud? After all, the cloud holds your data, shouldn't it be able to support your existing IT policies?
- How will you handle load balancing? Will you need local and possibly global load balancing that can be dynamically updated to include the new workloads you have bursted into the cloud?
- How will you charge back? Does your cloud bursting solution make it easy to charge back internal and external customers and set spend limits, controlling cloud sprawl and avoiding the auto-ballooning of cloud costs?
Whether you are cloud bursting or busting, as the great Lil' Wayne eloquently put it, "Make it Rain." Optimize your existing workloads and select the right provider - one that cannot only help burst your workloads onto enterprise-class cloud platforms, but also help you develop the IT strategies you need to grow your business.
Aditya Joglekar is director of cloud business solutions for Savvis, a CenturyLink company.