markrubens - Fotolia

Users still seek to resolve the public vs. private cloud cost debate

IT pros use everything from common sense to new tools to determine the best deal for running workloads in their own data centers or in the public cloud.

Price-conscious grocery shoppers compare unit prices to figure out the best value among ketchup brands.

If only it was that easy to calculate the cost of running a workload in the public vs. private cloud.

Part of the cost analysis includes how moving to the cloud can cut big-ticket expenses, such as closing down a data center, said Steve Nassif, senior manager for cloud services at Datalink Corp., based in Eden Prairie, Minn. If a company's data center is in the basement of a building where it could be located for the next 10 to 15 years, that analysis may come out differently, he said.

Often, Nassif's clients are hard-pressed to show the benefits, instead relying on promises from cloud providers that it will be cheaper in the cloud, with access to all the services they want. A few "have learned their lesson," he said, looking at their bills and figuring out the right things that can or should go to the public cloud. "[They] haven't pulled out full-scale."

Meanwhile, enterprises may want to leave a VMware environment, but IT pros see the Amazon Web Services (AWS) cost structure as unpredictable, and often find the math with private cloud simpler, said Ronen Kofman, vice product of product management at Stratoscale Ltd., which provides a software-only hyper-converged product.

"We have seen cases where customers have gone to Amazon, and then get on the phone to ask us to help them take some of it back because they can't control their costs," Kofman said.

Others may be compelled to go to the public cloud "for agility, flexibility and some of the PaaS [platform as a service] capabilities that are built in, plus auto-scaling or elastic scaling," said Sheraz Khan, manager of cloud engineering and services at food service provider Sysco Corp., in Houston.   

Yet another factor is staff time: There is a learning curve with both AWS and IBM SoftLayer, for example, so many enterprises have staff members with virtualization and VMware hypervisor skills, Khan said.

Khan said he uses an internal model and has determined the public cloud is cheaper for certain workloads, including a business object application running on 54 servers.

"You have elasticity at some level, but you are always overbuying and oversubscribing compared to what you would do in a public cloud environment," he said. "With public cloud, you get an itemized bill and know what you are paying for to the penny."

Translating cloud costs to Capex

The latest effort to distill cloud cost comparison metrics into a unit pricing model is a currency-based scheduler called Open Data Center Runtime Environment (OpenDCRE) from Vapor IO Inc., based in Austin, Texas.

OpenDCRE, an open source analytics engine integrated into Mesosphere's Datacenter Operating System, creates a way to describe the data center to applications. It pulls data from IT infrastructure and operational technology in the racks, such as power and cooling, along with metrics from the logical infrastructure, such as machine size, load size and resources, including Azure and AWS. Then, it applies machine learning to boil it all down to a uniform measurement: performance per watt, per currency.

This agnostic, application-based performance metric "translates to Capex and Opex" to help IT pros evaluate where an application workload would best be located: in the public cloud, colocation or on premises, said CEO Cole Crawford, who is also a co-founder of OpenStack and the founding executive director of the Open Compute Project.

Such a single metric that folds in IT and operational technology costs is valuable because it allows users to equally compare options, and there are few tools to arrive at such a calculation, said Kelly Quinn, data center trends and strategies research manager at IDC.

"It's frightening, but many are using Excel," she said.

Even so, tools won't eliminate the need for human analysis in cloud cost strategies for the foreseeable future. A tool can compare costs, but assigning value "can only be determined by human intervention," she said. Hunt's ketchup may be cheaper, but your cookout guests may prefer the taste of Heinz, and keeping them happy may be an important factor that's not measured in a cost calculation tool.

And for some enterprises, the determining factor in going with public vs. private cloud may not be cost at all. Many enterprises, such as healthcare, financial services and law firms, still don't have a choice, Kofman noted, because there may be compliance requirements to keep the data on premises.

Services available from cloud providers also draw some users, he said. Among attractive services from cloud providers are AWS Push Notification Service, AWS Simple Queue Service, databases, elastic load balancing and Amazon Elastic MapReduce.

"Still, cost control is a very big issue when they go out to the public cloud," he said.

Robert Gates covers data centers, data center strategies, server technologies, converged and hyper-converged infrastructure and open source operating systems for SearchDataCenter. Follow him on Twitter @RBGatesTT or email him at [email protected].

Next Steps

Public cloud providers offer new pricing models

Avoid these open source cloud hidden costs

Ten steps to control public cloud costs

Dig Deeper on Cloud automation and orchestration