Cloud Outages

How to recover after a cloud computing misstep

Early adopters share their lessons learned on ramping up, scaling back and avoiding disasters in the cloud.

1 2 3 4 5 Page 3
Page 3 of 5

Business Pressure: Why Can't We Do It Cheaper?

Many IT departments have learned to temper their business units' lofty expectations of the cloud with a dose of reality. In the fall of 2012, business end users at Aetna got swept up in the consumerization of IT trend and started asking IT a lot of questions about why they couldn't use software through a large hosting provider and just use cheap desktop PCs for $10 a month. To appease the business people, the IT team, along with an independent consulting firm, investigated.

"We initially set out to put desktop as a service in with a cloud provider -- but we very quickly learned that in our industry it can't be done in a secure manner while making it cost-effective," says Alan Pawlak, head of client services at the Hartford-based insurer.

"What a corporation puts into its [desktop clients] is way more than just the actual virtual machines," he explains. "A generic hosted virtual machine -- in the cloud -- it's still cheaper. But that means it's on a shared platform -- it gets upgraded with patches whenever they decide, with no consistency; and in that shared platform, nobody will sign on to the liability that you need to secure it the way you need to. When you add all these things up, you end up rebuilding what you have internally at a third party and it doesn't become cost-effective. It's simply not a solid solution."

Pawlak also notes network latency creates a very poor user experience in such a setup. "You're talking about very robust machines that run a full protocol stack back to your company to gain access to internal apps that you could never put [off premises]. That's where we stopped it," he says, adding that he walked his business colleagues through his findings.

Pawlak says he told business leaders that "the reality is, the industry really isn't there yet. We need to mature how data is stored, how applications are accessed and modernized before we can really get to that place."

Even in simplistic cases where it was an apparent no-brainer to move certain machines to virtualized clients, the cloud still didn't make sense because of the complexity of securing systems and providing access to applications from the cloud service into the internal environment. "It created latency issues, security problems, network problems for predictability and quality of service," says Pawlak. "We did not feel it was appropriate for a company in our industry to go down that path."

Instead, Aetna has decided to use the cloud just for certain tools that employees use every day -- like social collaboration tools and office software. "Those have a really good place in the cloud because I can encrypt the data, protect it and authenticate the access -- and there's not a real hook into the versions of the applications," says Pawlak.

Give and Take

Choosing which functions to move into the cloud takes discipline and compromise. While businesses like the quick time to market that cloud services provide, you need to avoid customization to keep costs under control, according to Oliver Bussmann, group CIO at UBS in Zurich. "We saw a lot of [cloud] implementation programs [with] 30% to 50% customization levels, and then the whole implementation effort goes up significantly," he says. "If you want to highly customize, that's the wrong way to use the cloud."

To get the full benefits of a cloud-based system, "you have to stick to the standard features that are offered because the cloud provider always likes to leverage the economies of scale," Bussmann says. "My experience is that maybe you get about 5% more functionality. So you need that discipline to stick to the standard functionality." In return, users get regular upgrades and new functionality every two to three months. "That mindset is critical for the success of the implementation," he says.

UBS uses a private cloud for support functions like human resources and service management, while customer data is kept on premises.

1 2 3 4 5 Page 3
Page 3 of 5
It’s time to break the ChatGPT habit
Shop Tech Products at Amazon