Maksim Kabakou - Fotolia
Put away your abacus -- a simpler calculation may help determine the most cost-conscious place for enterprises...
to do cloud computing.
For years, IT pros have asserted that the public cloud is lower than private cloud costs, but there has been very little hard data available to support that claim, mostly based on vendor-supplied information.
"There were all these myths flying around about the public cloud becoming a commodity," said Owen Rogers, research director in the digital economics unit at 451 Research.
The debate, he said, boils down to two key factors that most influence private cloud cost: labor efficiency and utilization of the private cloud. When not well used and managed, an OpenStack private cloud will cost more than the public cloud, but at a certain threshold -- specifically, according to his recent study, 400 virtual machines (VM) per engineer -- OpenStack costs become lower versus the public cloud. Conversely, when utilization and labor efficiency are low, the public cloud makes more financial sense over an OpenStack distribution.
"If you have engineers that are really good at running your private cloud and the cloud is highly utilized over its lifetime, in those cases, it is cheaper than public cloud," Rogers said.
The calculation is based on a basket-of-goods approach running similar workloads of compute, block and object storage on an enterprise private cloud, while also sending out that basket of goods to service providers and asking how much it would it cost. The study claims to cover about 90% of the infrastructure as a service (IaaS) market worldwide.
Owen Rogersresearch director, 451 Research
At the center of the price comparison is labor cost because an OpenStack engineer is paid on average $40,000 more annually than an engineer certified for VMware or Microsoft. The OpenStack Foundation has certifications that will help more OpenStack engineers enter the market, so perhaps that cost difference will change, he said.
Finding the most cost-effective place for a workload is a question often confronted by Robert Levally, chief technical architect at AlixPartners UK LLP based in London.
His cost analysis often involves a per gigabyte (GB) analysis of cloud services versus on-premises using reference storage area network architecture with a blend of flash and traditional disk to get a per-GB average and then compare it to a cloud vendor's per-GB cost.
The company's electronic discovery work often involves handling terabytes or petabytes of data from around the world, and in some cases it is impossible to ingest all the data and move it to cloud. In that case, public cloud costs become largely moot.
If an application has known characteristics and is in a steady state, it is more expensive -- up to three times as much -- to run it in the public cloud, said Chris Cicotte, portfolio marketing director for cloud and a principal systems engineer at Dell EMC, whose past experiences include running a data center for a Fortune 5 company. In three to five years, that type of application will be cheaper and enterprises will have more control over it in their own data center.
The cost analysis is much different for a company that doesn't have its own data center or facilities, he said.
"We realize that the use of public cloud is strategic in your environment and companies aren't shying away from it," Cicotte said during a presentation at the recent Dell EMC World event.
A cloud-first push within an enterprise may or may not make sense, depending on the level of restrictions, or if it means moving workloads to only one provider or moving every workload, Rogers said. Large enterprises should always keep options open because there might be a different provider out there to meet all of an enterprise's needs at a lower price.
Small and medium-sized enterprises, meanwhile, may not have the scale to make private cloud cost-effective, lack the needed utilization and labor efficiency or the appetite to hunt around for different cloud providers. For those evaluating an OpenStack distribution from a vendor versus raw, the question is answered if the distribution can save 5% of manpower costs in managing the private cloud, Rogers said. OpenStack distributions help improve management and installation and provide a "clear advantage over using out-of-the-box OpenStack" for most users, he said.
The mighty dollar doesn't rule all
The study didn't look at some of the other equally important factors that draw enterprises to a private cloud, including the fact that it is fully managed in-house and they know where the data is located.
A workload running on-premises can guarantee performance versus the public cloud, where it becomes the function of an enterprise's best effort unless you pay more, said AlixPartners' Levally.
His company doesn't write much software and typically runs third-party applications and does little coding, so in many cases it does not make sense to move those applications to the cloud, he said. Moreover, it is often more expensive and risky and his company doesn't have total control.
Even running traditional software presents a dilemma when encountering regulation and client requirements. A customer may insist that data stay in a certain geography -- France, for example -- where the company doesn't have a data center, which means turning to a local cloud provider.
"Cost is still a factor because, just to do business there, if that drives the cost up too much, then the client walks away," Levally 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].
How to save private cloud costs before and after deployment
Solve OpenStack deployment issues
Enterprise workloads for cloud increase