The major benefits of cloud computing are the abilities to scale quickly and deploy additional resources as the need arises. But many businesses choose not to move their infrastructure lock, stock, and barrel into the cloud, instead preferring to keep core components in-house. Nevertheless, those companies can still take advantage of cloud scalability using a process known as cloud bursting. Cloud bursting is the augmentation of existing in-house hardware with cloud hardware in response to increased demand.
Let’s take a very simple, and artificial, example. You run a relatively popular eCommerce store that is hosted on in-house servers. Coming up to the holiday season, you’re aware that traffic may well exceed the capacity available, which will result in a poor experience for shoppers. It’s not economical to buy extra servers, because for most of the year it will lie idle. In order to accommodate the additional traffic, you can plan to “burst” into the cloud — when traffic exceeds the resources of the in-house infrastructure, new requests are routed to a synced copy of the site in the cloud. The cloud portion of what is now a hybrid cloud can be scaled according to demand. As traffic hits its peak, more cloud servers can be brought online, and then they can be spun down as demand wanes.
In reality, cloud bursting can be quite complex, keeping the in-house and cloud sites synchronized is not easy. For example, if you are running databases both in-house and in the cloud, there is considerable complexity involved in keeping data consistent between sites. But it’s not an unconquerable difficulty as Dotan Horovits has explained. After all, the problem of load balancing databases is well understood, the only major difference is accounting for potential latencies introduced by the addition of a cloud component.
Web hosting isn’t the only scenario in which cloud bursting can be useful. Whenever a company needs to process workloads that exceed their usual capacity, there is a case for moving a portion of the workload into the cloud. We’ve talked about bio-infomatics computing here before; bio-informatics workloads like DNA sequencing involve the periodic processing of huge amounts of data to produce results that use exponentially less storage space. By bursting the bulk of the sequencing and analysis workloads onto the cloud, and using in-house infrastructure for the storage and manipulation of the results of those analyses, organizations can reduce capital expenditure. Many businesses have comparable periodic resource demands.
Cloud bursting allows organizations to leverage existing infrastructure while taking full advantage of the cloud’s unique abilities.