Like a hurricane barreling up the coast, cloud computing is gathering strength and expected to make landfall shortly. In some IT shops, cloud is already here, bringing the winds of change along with it.
But unlike a natural disaster, cloud computing doesn’t destroy everything in its wake – merely transforms it.
In the data center, IT operations managers who have experimented with cloud — private, public, and otherwise — report it’s already had a profound effect on the way they run their IT environments and do their jobs. And they expect even more changes to come.
What do IT operations professionals that have felt cloud’s effects have to say to the uninitiated? Brace yourself.
Cloud has changed the conversation about capacity-planning.
Cloud computing shrinks IT staff, alters operations
Even tentative steps toward cloud computing have a way of snowballing into something much larger and unexpected.
At the most basic level, cloud technologies can reduce the number of administrators needed to manage an environment.
That’s what happened at Walz Group Inc., a provider of document management services in Temecula, Calif. In 2010, the company implemented a FlexPod private cloud based on Cisco Systems Inc.’s Unified Computing System (UCS), VMware Inc. virtualization and NetApp Inc. storage, plus cloud management software from Cloupia Inc.
In short order, the IT staff dropped from six to two, and its server-to-admin ratio from 15:1 to 50:1.
“We had been having trouble scaling out, but when we deployed the converged infrastructure, we were able to manage it more effectively,” said Bart Falzarano, chief information security officer at the firm.
Others say implementing a private cloud has helped IT operations stay abreast of demand but doesn’t go far enough.
“Before, there was an approval process, and we would either say yes or no,” said Curtis Gunderson, director for virtualization architecture at Unum, a Fortune 500 insurance provider in Chattanooga, Tenn. “Now, rather than go through the approval process, we basically say yes to everything.”
Unum’s private cloud includes 1,400 VMware virtual machines (VM), plus another 400 physical servers, and an array of capacity management, cloud automation, orchestration and self-service provisioning software.
The firm would also like to start using external cloud resources as early as next year to keep up with the demand to deliver services, Gunderson said.
The availability of endless cloud resources is very appealing to enterprise shops that wish to get off the capacity-planning treadmill.
“Cloud has changed the conversation about capacity-planning,” said Ian Rae, CEO at CloudOps, a managed services provider that offers private and public cloud management and integration.
The classic example of this is Zynga Inc., the online game maker, which revs up new games on Amazon and brings them back in-house once Zynga has a better handle on its performance needs.
Compared with the traditional world of static IT operations, where “the only way to get more capacity is to sign a P.O.” — the cloud allows you to provision, re-provision and de-provision at will, coding to your infrastructure using a whole new class of automation and orchestration tools.
“You can take more risks, be more creative and dynamic,” Rae said. “You no longer have to over-invest in a project that doesn’t end up doing very well.”
Bursting the cloud bubble
That’s the vision. The reality of using public cloud for enterprise IT is still far away.
Cloud bursting, for instance, still leaves a lot to be desired. Right now, “it’s more of a conversion,” Falzarano said, than a seamless migration. The company is eyeing emerging physical-to-cloud (P2C) and virtual-to-cloud (V2C) players such as RiverMeadow Inc., whose enCloud cloud management software promises to ensure migration between private and public cloud stacks, including OpenCloud, Cloud.com and Amazon Web Services.
Nor is there a lot of clarity about cloud security. “We work with a lot of highly regulated companies, and their requirements get passed on to us,” Falzarano said. To outsource even part of its operations to the cloud, “we would have to be assured of our security and infosec requirements,” he said.
The lack of maturity of the application development and platform as a service (PaaS) market holds Unum back, said Gunderson. “When it comes to hybrid cloud and bursting, we are still looking around,” he said. “We want to be able to take a workload and take it down and shift it around dynamically without an outage. Today, we can’t do that.”
Unum is doing a proof of concept using Ruby and VMforce, but that doesn’t jibe with its Microsoft.Net-centric developers. Currently, Unum has to choose between its preferred PaaS platform and its application development environment.
“On our roadmap, we wouldn’t have to choose,” said Gunderson.
Parsing the operations stack
IT operations professionals know that despite these obstacles, public cloud will infiltrate their environments. And when that happens, things are going to change.
The IT operations team at Unum has traditionally been responsible for three broad areas: rack-and-stack hosting concerns, the operating system layer and the application integration layer, Gunderson said.
“I fully expect the first two are going to go away,” Gunderson said. “The hosting layer is nothing that we should worry about,” and with PaaS, the operating system layer will become “more agnostic, more cookie-cutter,” he said.
At the application layer, meanwhile, IT will act more as consultants involved in platform selection and troubleshooting, Gunderson predicted.
Four of the six admins that used to manage the infrastructure estate at Walz Group have moved “up the stack” and into the application development group, Falzarano said.
Falzarano, meanwhile, spends a lot of his time evaluating workloads and their architectural, performance and compliance requirements, and mapping that against the vast array of available cloud services.
“I feel like this is where my role has really transitioned,” he said. “Three years ago I wasn’t really focused on cloud or SaaS. Now I’ve become a kind of cloud broker agent.”
This is part of a broader cultural shift in IT operations, said James Urquhart, vice president of product strategy at enStratus, a cloud management and governance provider in Minneapolis.
“In an app-centric operations model, we can’t have the silos we used to have: servers, networks and storage,” Urquhart said. “Operations tasks are breaking down along different lines.”
Today, the ops team tends to provide the entire stack, from facilities and hardware “all the way up through the operating system and middleware,” Urquhart said.
In a cloud model, he expects two classes of operations teams: infrastructure operations that stop short of the operating system and middleware, and a service operation teams focused on applications and services.
Salesforce.com, for example, has people that deliver it, the infrastructure ops team, and people that consume it. But even the people who consume Salesforce still have to do a certain level of operations, he said.
The canonical example of this idea is the Netflix Chaos Monkey, where the company’s engineers randomly kill instances and services running on the Amazon Web Services (AWS) environment, in order to uncover failure points or poor performance.
“They’re not looking for network outages, but application outages,” Urquhart said. Although Netflix may not have many servers anymore, its engineers still have plenty of opportunities to deliver applications.
Wanted: Ops guys who code
Traditional IT operations professionals that want to boost their career in the cloud era have to learn how to code, said Jared Wray, CTO at Tier 3 Inc., an enterprise cloud hosting provider based in Bellevue, Wash.
“You may hate it, but it’s going to make you so much more valuable,” he said.
Already, Tier 3 has stopped hiring any administrators that don’t have some development experience.
“We only hire DevOps guys here,” Wray said. “You have to be able to code to work on our team.”
That’s in contrast to traditional IT departments, where developers and operations are run separately. But in cloud environments, “DevOps is huge. The apps guys are now part of the ops team and vice versa.”
Learning to code doesn’t necessarily mean taking a class in Java or C++, Wray said. In fact, that may be counterproductive. Instead, he recommends downloading an open-source PaaS platform like CloudFoundry, spinning up an application on it, and learning to code against it with Ruby.
Others recommend learning other platforms or languages, but the point remains: IT operations folks have to move out of their comfort zone to stay in the game.
“Traditional forms of operations are increasingly being offered as a service,” said Urquhart. For people who have made their living in the data center, learning these new cloud skills and platforms “is a question of staying on the side of growth.”
Otherwise, “you’ll become less and less relevant.”