Nigel Beighton, VP of technology of hosting company Rackspace, outlined the five areas that firms should focus on in order to get the most out of Infrastructure as a Service (IaaS).
Speaking at Structure Europe in London, Beighton said that while more and more companies are now using IaaS, it is only once they learn to properly manage the main advantage it brings - scalability - that they can truly benefit from the cloud.
Further reading Rackspace public cloud struggles against Amazon's AWS RackSpace to provide OpenStack hybrid cloud to CERN Rackspace CTO: Oracle and SAP are starting to panic
Previously employed as CTO at booking site lastminute.com, Beighton reminisced about the days when they would bring in masses of extra machinery and expertise every time they wanted to scale up to meet a peak in demand.
"I'd spend £100,000 on new servers and contractors, all to service a peak that lasted 16 hours," he said. "The rest of the time we'd just play Doom and Quake on the servers. Contrast that with today when you can get extra capacity to deal with a spike in demand for £200."
The scale of this cost differential, Beighton said, is fuelling the cloud revolution, but he claimed that organisations are adopting IaaS without really understanding it.
The five essential toolsets outlined by Beighton to make the most out of IaaS, are as follows:
API management
"If all you are offered by a cloud provider is a single UI to click to deploy your servers, walk away," he advised. "Unless you have more control than that you will not be able to scale up and down fast enough to reap the benefits."
Automation
There are plenty of tools available on the market, said Beighton, that will automate the scaling process, such as RightScale and Salt Stack. These are essential if you are to get the most out of IaaS.
Monitoring
"We live in a cattle world not a pets world," said Beighton.
"If one server goes down we don't need to care about it too much, but we do need to know where in the cloud application bottlenecks are occurring, which is more complex than on a traditional network.
"This means that you need to optimise applications for the cloud infrastructure so they can scale and self-heal," he said, insisting that some sort of cloud application performance management system is essential for optimising IaaS.
"I don't worry how may servers it takes to run my application, but I do worry about how it scales," he said.
Security and control
Most cloud deployments are on a shared environment, said Beighton.
"Therefore you have to think about security because it's less in your control. Security is typically a static thing, but now you're having to think about what it means when you scale up and down, whether it's encryption of the data, or controlling access in some way. The toolsets you need are typically different from the static toolsets you have now."
Watch the costs
Since many cloud services are billed on a pay-as-you-go basis, it is important to monitor what effect scaling up and down has on the overall costs before the bill comes in, said Beighton. Understanding and linking your cloud to some sort of cost tracking system is necessary if you are to avoid a nasty shock, he concluded.
