Be prepared to duck, because some data may return to your enterprise data center as workloads return to on premises...
from cloud computing -- the so-called boomerang effect.
Nearly two out of three CIOs (62%) in a recent Pacific Crest Securities survey, 2017 Cloud and Infrastructure Priorities, have brought workloads back to their data centers, rekindling the on-premises vs. cloud computing debate and suggesting the cloud may not be the answer for everything.
Survey respondents came mostly from technology, financial services, manufacturing, public sector and retail businesses and two thirds have an IT budget of $50 million or larger.
Many enterprises have been unpleasantly surprised when their monthly bills from public cloud providers show how much they actually spend, said Scott Lowe, an independent technology analyst who spoke at a recent Data Center Reset event hosted by TechTarget.
Nasdaq Inc. in New York, for example, underwent a cloud computing push a few years ago but discovered that, in some cases, it cost more than keeping data on local servers, he said. The American stock exchange has begun to move some workloads back to private data centers. The company will continue to use a combination of public and private clouds to run given workloads based on costs and application demands.
The boomerang effect usually occurs with technology companies and larger enterprises, according to the Pacific Crest survey.
Public cloud is especially appealing to startups with uncertain models and the ability to scale. "To be able to spin something up in hours and have it running, and your code deployed just working and scale up, is amazing," said Andrew Hamilton, CTO at Parkmobile LLC in Atlanta. The company handles parking payments for on-street spaces, parking lots and garages in communities across the country, including in Atlanta, Washington, D.C., and Los Angeles.
Start-up costs for companies, such as hardware, Microsoft licenses and engineers, can total tens of thousands of dollars on day one, but they can be largely avoided by leveraging cloud computing, said Hamilton, a 20-year technology veteran. However, as a company and its needs mature, it can start to move away from the cloud.
Parkmobile meets compliance requirements with workloads that run in its own hardware and firewalls, isolated to its environment.
"That is easier than saying it is the public cloud, and you have to trust and believe that Amazon doesn't have access to any of the things we are putting out there," he said.
Aaron RalloCEO, TSO Logic
About a quarter of Parkmobile's business relies heavily on Amazon Web Services (AWS), through its reservations business for events such as Atlanta Braves and Atlanta Falcons games.
TSO Logic grappled with where to put workloads in 2015 and developed software for its customers that has no allegiance to an enterprise data center, colocation facility or cloud computing.
"In every environment we go into, we find things that would be more expensive to operate in the cloud every single time," said CEO Aaron Rallo.
The company's analysis of 25,000 customer instances that run in an enterprise data center found that 45% of virtualized instances could run more economically in the cloud if migrated as is, and that would reduce costs by 43% annually.
No enterprise has come to TSO Logic looking to exit the cloud, Rallo said. A more representative example, he said, is one video-streaming customer that has several data centers but still must scale rapidly when it streams NFL Sunday ticket and the NCAA Final Four.
The company typically gets workloads up and running through infrastructure as a service (IaaS). Once initial deliverables are met, the company backfills through its own on-premises infrastructure because the cost of operating on premises is less expensive, Rallo said.
For the past few years, Parkmobile's Hamilton has weighed the pros and cons of on-premises vs. cloud computing deployments. He compared the costs a year ago and found that cloud computing wasn't a significant money saver. But he said he will never return to his company's traditional data center, because it's unnecessary for companies that specialize in software -- focusing on the product is viewed as more important than worrying about the power and cooling needed to run a data center.
One of the initial attractions of cloud computing was that it would mitigate large capital expenses and allow compute to be bought as smaller chunks of operational expenses, Lowe said. But there will always be circumstances where cloud computing will not save money in the long term, 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 firstname.lastname@example.org.
Three signs your app will cost more in the cloud
Use autoscaling, tagging to save costs in cloud
Five tips to optimize cloud costs in 2017