By Anand Akela on Jan 03, 2013
Contributed by Eric Tran-Le, VP, Product Management, Oracle Enterprise Manager
With public clouds users can now compare compute pricing pretty much like they would compare a car models. One could argue that compute power and services are different than a car but the fact is that users can compare costs.
Nothing has changed and everything has changed
Cost wise nothing has changed in a sense that the decision to implement a private cloud versus hosting on a public cloud is driven by the same factors than outsourcing or on-premise. It goes well beyond simple cost comparisons to touch upon key principles such as:- Control of costs
- Visibility of costs
- Fairness of costs
Everything has changed in a sense that there is now a cost “predictability” expectation from users derived from a notion of “unit of compute” (the equivalent of standard energy Unit of Work) from which you can predict your resources consumption and infer total costs (a 45mpg rating means that you can drive 450 miles with 10 gallons or $40 if $4 per gallon). This is a more fundamental change in a sense that IT needs to work with the finance department to rationalize the cost-of-compute(Infrastructure + SW platform) and link it to the cost-to-serve (incident, problem, change & configuration management), the end goal is to produce a standard unit of compute that can be applied to various products and services configurations.
In part 1 of this blog, we will highlight how a cloud chargeback system can help addressing the key principles mentioned above and in part 2 we talk about how to create a profitable cost recovery model.
Key Principles of Chargeback System
Control of Costs
is not defined cannot be controlled”, “What is not controlled cannot be
measured”, “What is not measured cannot be charged”.
A cloud chargeback system provides a framework by which you can define a standard unit of cost for products and services. You can set unit of costs for resources based on hardware tiers or database high availability tiers, you can add fixed costs for on-going support costs or for facilities driven cost such as real-estate, power, HVAC,… but more importantly you can organized them by business units and cost centers so that you do have a cost structure hierarchy to report in aggregate or by business units. To define the right set of configurations, you need to baseline the current workloads utilization, characterize the patterns group of users and type of workloads and since you will use based compute metrics (CPU, memory, I/O,…) to measure utilization it is important for you to define which layers of computing ( Servers, Middleware, Databases,…) will reflect fairly the actual usage pattern of an application.
Visibility of Costs
Among all the features showing business units their VM consumption and allocations is the one that will start to change the relationship between IT and users more fundamentally. Actual utilization rate, historical trending, heat map displaying under and over utilized servers but also virtual environment configuration attributes such as high availability and regulatory compliance reports will contribute to a better cost-to-serve transparency. Another important capability to look for is how you can identify servers that cannot be part of a shared pool of resources due either to legacy applications or simply very high workload requiring dedicated machine this will help when you start sharing with the user the calculation of his charge plan.
Fairness of Costs
This the ONE difference with traditional chargeback models in a sense that metrics collected for charge calculation are actually “IT compute metrics” on top of which you add “services fee-based” metrics. In traditional chargeback direct or indirect model, the finance department tends to use non IT metrics such as # of users in a business unit as a % of aggregate IT costs or data center space taken by rack of servers,…this has led to a lot to the lack of transparency and the sense of “unfair” chargeback. As part of the charge plan calculation the characterization of fixed versus variable costs is essential to calculate the unit of charge.
Remember that in the cloud, users do not want to pay for resources they do not use and there is an expectation that they can dynamically request more resources, allocate and de-allocate resources. A cloud chargeback system will help you transition to a usage-based pricing of your IT resources will the key principles of control, visibility and fairness of costs. Last but not least, you do not need to go for the big-bang and turn chargeback right away you can implement showback and specific chargeback initiatives so that the business units can work with you on an efficient chargeback model.