As contentious as IT chargeback can be, the model is fundamental to cloud computing. The central idea is that computing resources and services are metered like electricity, so
But what metrics can IT departments use to charge back business units for cloud services -- or for any service, for that matter? It's a complex question that grows even more so as you move up the cloud computing stack, according to experts. At the Infrastructure as a Service (IaaS) level, you could charge for CPU cycles, I/O, bandwidth, memory and disc space; if Platform as a Service (PaaS) is the concern, add application development and support. At the high end, a Software as a Service (SaaS) metric should evaluate the business value of the service -- the number of customer requests responded to within a given period of time, for example.
There are tools that help IT departments gather data about CPU cycles and so forth, but the added support is a soft number that can give business units pause; moreover, the value-added SaaS metrics can be tough for IT staff to wrap their heads around. Add to those the cloud's virtual, shared infrastructure and its elasticity -- accommodating resource spikes in various workloads -- and it can be hard to know whom to bill.
Still, IT departments need to figure out chargeback, and the sooner the better, to compete with the likes of Amazon.com Inc.'s Elastic Compute Cloud, or EC2, which business units can easily bypass their IT department to buy, then still expect internal IT to support.
"Chargebacks have been a bone of contention forever, and the advent of cloud services really forces the issue," said Julio Gómez, co-founder of Innovation Councils LLC, a Concord, Mass. -based company that brings together CIOs in various industries to discuss emerging technologies. "Third-party pricing is relatively transparent and the services are readily identifiable, especially in IaaS, so internal customers can look at what they pay for servers, compute time, storage, etc., and compare it to what Amazon charges," he said. "It puts IT in an uncomfortable position."
Complicated metrics part of IT chargeback
In the traditional chargeback model, an IT department might divide its budget for services by the total number of business units it serves. Of course, this gets problematic when one business unit thinks it's getting ripped off. In the cloud, that scenario gets even more complex, because IT needs to consider the rate and time of consumption.
Consider the bill for a typical utility, such as water from a faucet or electricity from an outlet. With water, you would be charged for what you use, whether you turn the faucet on full blast or let it drip for 10 hours. From an IT perspective, supporting a high I/O for a short time might require a different architecture than a consistent usage pattern would -- and likewise, a different chargeback model. And, like an electric meter that spins faster as more devices are plugged in, the IT chargeback model needs a layer of granularity to reflect that, according to Daniel Weiss, a former cloud strategist at Unisys Corp.
In a virtual world where systems are combined, not only is the metering more difficult but the possibility of a system failure looms as well. Virtual servers frequently are oversubscribed, meaning that they probably would crash if all the clients on them required maximum compute power. So, what do you do when one application goes overboard? Experts suggest charging an overuse fee.
The point is not to turn IT into bean counters, but to make the business units more aware of the resources they're using and to demonstrate that IT investments are supporting strategic initiatives, according to Weiss. "Metrics are inevitable," he said. "I would even say they are now mandatory."
All talk, no IT chargeback
Because of the complexities of IT chargeback, not many enterprises have implemented it. "A lot of people talk about doing it, but it takes something to do it well and actually look at the data," Weiss said. A business might have software to measure IaaS components, he said, "but it's rare that companies fully implement the software -- usually, 20%."
Internal customers can look at what they pay for servers, compute time, storage, etc., and compare it to what Amazon charges. It puts IT in an uncomfortable position.
Julio Gómez, co-founder, Innovation Councils LLC
In addition, the software programs out there don't go far enough, according to experts, who say the self-service aspect of cloud computing needs to be combined with metering through an automated portal. VMware Inc. is said to be working on such a product in a bid to "own" the private cloud stack, according to experts.
Unfortunately, such tools weren't available when the cloud arrived, and there are no standards for IT chargeback with self-service automation -- a situation vendors hope to change by starting a council within one of the standards bodies, according to David Link, president and CEO of ScienceLogic LLC in Reston, Va., which markets a customizable cloud measurement and provisioning portal.
Link has been talking about the council with Intel Corp., which is building a private cloud on 20,000 servers to establish best practices for enterprises. The Santa Clara, Calif.-based company recently released a white paper describing its custom-built portal for managing metering and self-service. "Later this year, we plan to extend use of the portal to all private cloud development and test environments, and then to establish the portal as the standard interface for requesting IaaS capacity for production applications," the document states.
Clearly, if Intel is just figuring this out, enterprises can't be blamed for dancing around the IT chargeback issue. "It's one of those mythical, last-mile IT items that never seems to get completed," Link said.
"But it's fundamental if you're talking about a cloud delivery model," Weiss said. "The only way for a company to provide services in a meaningful way is to have a chargeback system."
Let us know what you think about the story; email Laura Smith, Features Writer.