Cloud bursting is a way of using public cloud resources to deal with spikes in demand. When it comes to solutions, Nick Hardiman asks: Why trade capacity problems for complexity problems?
What is cloud bursting?
The name cloud bursting comes from the telecoms world — cloud and bursting are both old networking terms remixed by new cloud vendors. Telecoms providers have offered their customers extra burst capacity on their MANs and WANs for many years.
Cloud bursting is a way of dealing with peaks in IT demand that can't be handled by on-premise hardware. This technique is useful to organizations with on-premise private cloud, rather than the co-located kind. It requires the kind of hybrid cloud formed by extending the enterprise network from on-premise to off-premise cloud services.
The promise of cloud bursting is to do a better job of fitting capacity to demand. Organizations can't squander cash on a kit that spends half of its time idle (if they could, no one would have bothered replacing their internal telephone networks with VoIP). It would be so much easier to cope with variable computing demand by maintaining only enough hardware for predictable loads and shipping the rest — the spikes in demand — to the cloud. The process of deciding how much hardware to buy for next year would rely more on metrics and less on gambling.
Why making cloud bursting work is difficult
Cloud bursting is proving hard to get right. Making cloud bursting work is difficult for many reasons: systems can't talk to each other, distributed applications require re-architecting, and internet traffic must be secured. If the cloud bursting solution requires handcrafting by developers and administrators, taxes the training budget, and strains the support teams, why bother? What's the point in trading capacity problems for complexity problems?
The company that can make cloud bursting easy is onto a good thing. Easy doesn't have to be a single cloud burst tick box — it just has to remove some of the pain of wedging incompatible systems together.
Cloud bursting solutions
Cloud vendors know how attractive the idea of cloud bursting is, and many solutions exist. RightScale, Eucalyptus, and OpenNebula provide tools for building application-level cloud bursting. Microsoft provides platform-level cloud bursting to Microsoft Azure.
Some cloud infrastructure management systems can talk to Amazon; for instance, VMware has Application Director, and Red Hat has CloudForms. These toolkits help to build a cloud bursting network, but leave much work undone.
What about VMware and Amazon Web Services (AWS)? VMware, king of private cloud, does not come with an "enable cloud bursting" checkbox for AWS, king of public cloud. HotLink, a company in California, sticks VMware and AWS together with its Hybrid Express product. Amazon resources appear in VMware's vCenter management console, which makes cloud bursting easier for an organization's VMware ops guys. Hybrid Express also plugs into Microsoft System Center.
The lifecycle of cloud bursting
The idea of cloud bursting appeared years ago, but it won't live up to its promise until tech staff are given better tools for building heterogeneous networks. The tools must help the tech staff by bringing simplicity to complexity, the way Hotlink Hybrid Express does for VMware and AWS.
Will cloud bursting still be awkward a few years down the line? Maybe the IEEE Intercloud standard will remove these barriers to creativity. Maybe Docker payloads will make virtualization redundant. Or maybe computers will disappear into the background in a big puff of IoT.
What do you think?
Does cloud bursting solve real capacity problems, or is it just a more complicated form of vendor lock-in? Tell us what you think in the discussion.