Consumption-based pricing and the ability to spin cloud servers up and down as needed have opened organizations up to a whole new way of doing business – one that requires a lot less upfront capital. But something that still eludes many organizations is how to monitor the costs of their cloud usage before it gets out of hand.
Lynn LeBlanc, CEO and founder of hybrid IT service provider Hotlink, says that public cloud is typically more of a so-called black hole than on-premise infrastructure. “One advantage of on-premise infrastructure is you buy it, and it’s yours, and it doesn’t cost you anymore to use it up,” she says.
“If you’re on a pay-as-you-go plan of public cloud, while you get that upfront benefit of not having to extend all that capital…really managing that consumption, it’s a new problem that people aren’t used to having to solve,” she tells Talkin’ Cloud in an interview.
To address this issue with its own corporate IT customers, HotLink launched managed services last year, in conjunction with its hybrid HotLink Cloud-Attach Platform, that provide AWS cost optimization and load optimization. These services were born out of the company’s own experience dealing with what LeBlanc calls “bill shock” at the end of each month.
“I would get these [AWS] bills at the end of the month and honestly I never knew how much they were going to be, and they weren’t really organized in a way that we could deconstruct them,” she says.
According to LeBlanc, while an Amazon bill shows her how much data was transferred to and from the cloud, how much storage she consume and other metrics, it “will not tell me who was doing what, what were they doing with it…so when it came to trying to figure out, ‘wow, suddenly we got this monster bill that, by the way, was way more than we budgeted for, why? What were we doing?’”
“I know that our engineering VP always dreaded when the bill came in because he knew there was going to be this fire drill to try and figure out what it was,” she says.
For engineering it was particularly problematic because it was impossible to figure out how many cloud resources individual developers were using. So the company developed its own tools to monitor cloud usage, some of which use the Amazon API.
“What we found was the reporting we created for ourselves could be really useful for our customers,” LeBlanc says. “We gave them some of those tools but then we found there wasn’t that much discipline; that you really have to have if you care about managing cost in the cloud.”
“There are a lot of tools that do all kinds of things related to cost but it still depends on somebody really actively managing it. We found that when it came to corporate IT, and particularly the upper mid-market, I just don’t think there was the discipline around 'you’ve got to watch this stuff every day' if you really want at the end of the month to have the economic benefit that you envisioned when you started to use this resource.”
HotLink started offering managed services around its various products last year, including disaster recovery as a service and general cloud management, but it wasn’t until August 2016 when the company productized its cloud cost optimization and load optimization and made it part of its managed services offerings.
HotLink monitors AWS cloud costs on behalf of its customers with its cloud cost optimization module. LeBlanc explains: “We find out from them their budget for public cloud usage for each month, we monitor it daily to see what’s powering on, what’s not, perhaps an extra-large instance is powered on but there’s almost no CPU usage. This means somebody should have just powered it off.”
For companies that are just getting started in monitoring cloud costs, LeBlanc says it can be hard to figure out how to budget. Similar to a household budget, it is best to have some number of months of usage before you can create a realistic budget. It also requires a bit of a behavioral change, she says.
“There’s a bit of corporate IT that’s sort of opaque. They don’t necessarily know what all these applications are doing so let’s say for example they’re using the public cloud for disaster recovery. They don’t necessarily know how often a given application goes through a major update cycle for their database. They just never paid attention to that because they don’t manage things at the application level for the most part; they’re looking at it more holistically.”
“Not being application aware from an operational point of view is a change in behavior that is necessary when you have a new unit of financial measure in the public cloud,” LeBlanc says. When you don’t know the application, there is a tendency to overprovision, she adds. HotLink makes recommendations to customers about right-sizing the pre-defined instances based on the performance and utilization of the resources, she says.
With its load optimization module, HotLink provides load balancing recommendations for on-premise data transfer; intelligent scheduling for bandwidth management; and AWS account configuration for fewer bottlenecks and faster throughput, the company says.
“[W]hat you’re trying to do is keep from saturating the network that is being used for a lot of purposes…we haven’t had any customers where they just didn’t just have enough bandwidth but if it’s not being accurately managed they will have bottlenecks,” she says.