It's one of the fundamental questions of cloud computing: Is it less expensive to run workloads in a public cloud than in an on-premises IT environment?
The answer, says Charlie Burns, a researcher at boutique advisory firm Saugatuck Technology, is "a definitive maybe."
The answer varies depending on individual customers and use cases. But, Burns found some general rules of thumb. "It comes down to the basic question of, 'How good are you at running your in-house environment?'" he says.
Large enterprises with highly optimised IT shops tailored to their business' needs may find cloud computing to be more expensive. But, if a company has workloads that ebb and flow in their use of compute power, then the cloud can yield substantial savings, Burns found.
"Cloud IT infrastructure offerings can yield significant cost reductions for various types and sizes of workloads. But the reality of such savings is highly dependent upon the type of workload, its suitability to cloud, and the level(s) of efficiency and optimisation of in-house IT resources and operations," Burns' report states.
The takeaway: Do your homework to find out what this means for you specific use case
The central theme of Saugatuck's newest research -- that optimised in-house IT is better for static operations and the cloud is a better deal for dynamic workloads -- is backed up by other major research firms as well.
In a report last year, Forrester said the pay-per-use model of public cloud computing alleviates customers from over-provisioning on-premises resources to handle peak demand in their own IT environments. "When your application load varies and you reserve capacity in advance, the cloud will likely always be a clear cost winner," Forrester analyst Dave Bartoletti wrote. The Forrester report has a caveat, though. "Cheaper isn't always better," it says; perceived concerns around security, performance and reliability have left some IT managers just more comfortable using their own in-house IT.
Burns, from Saugatuck, says there are a couple of important factors IT decision-makers should consider to help determine if cloud is a more cost-effective option. While most public cloud-computing resources are pay-per-use, many also come with pre-configured virtual machine image sizes. Amazon Web Services, for example, has dozens of options of VM sizes that can be spun up and down by the hour. But even with a broad choice of VM sizes, there may not be one that is optimised for the workload that is being run in the cloud.
There are some solutions to this: An expanding roster of third-party tools have emerged that help customers optimise their clouds and ensure they are the right size for their applications. There has also been a rise of customisable cloud computing VM sizes, from providers like Dimension Data, ProfitBricks, CloudSigma and, to a lesser extent, Microsoft Windows Azure.
Then there's the issue of hidden costs. The cost of a public cloud service is not just the VM per hour price, or the per-gigabyte of storage cost. Data egress and network bandwidth expenses are "gotcha" costs, Burns says, that many customers overlook. In Amazon Web Services' cloud, for example, it's free to upload data to the cloud, but the company charges to take data out, and the network resources that go along with that. Basically, cloud providers want to make their services sticky.
Still, the cloud is a good deal for many workloads, says Mike Pearl, a consulting partner at PricewaterhouseCoopers, who leads the organisation's cloud unit. Having a cloud -- be it a public one or a behind-your-firewall private cloud -- will usually yield savings for a company because of the efficiencies gained with having on-demand, elastic resources. "Moving to a cloud architecture has demonstrable benefits around cost and efficiencies," Pearl says. "The bigger question is should that be built internally or in a public cloud? That's where more analysis is usually needed."
The overall point is highlighted in another report from Forrester by cloud expert James Staten, titled "Cloud computing is not the future of IT." In it, Staten argues that the cloud is essentially another tool in the IT administrator's toolbox; it's good for some use cases, but not ideal for others. The key is knowing when to use it for what, and choosing the right platform and provider. But cloud isn't replacing traditional IT: that will always be around.