The Smart Electricity Grid
New research makes possible time-variable pricing of cloud services based on variably-priced power from a smart grid
Cloud computing data centers often consume energy at the rate of megawatts, like other large industrial systems. Large consumers of electricity likewise often have to deal with time-varying pricing of electrical power from suppliers, who increasingly prefer “smart” (time-varying) metering and smart power grids. The cost of electrical power is often the largest component of the cost of cloud and other IT services.
So how can cloud computing servicesoffer variable tariffs to their consumers based on the variable pricing of power from a smart grid? This question is addressed in the research article “Power-Aware Cloud Metering” by Akshay Narayan and Shrisha Rao, appearing in the most recent issue of the IEEE Transactions on Services Computing (volume 7, number 3,), a research journal of the IEEE Computer Society.
Akshay Narayan, presently a Ph.D. student at the National University of Singapore did this work jointly with Shrisha Rao, a professor at IIIT Bangalore, as part of his Master of Technology (M.Tech.) thesis in information technology. In their work, Narayan and Rao arrive at a metering mechanism for cloud services in which the price of a cloud service tracks the variable input cost of electricity from a smart power grid. The power-aware cloud metering developed is a dynamic pricing and billing model where tariff for a cloud service is varied in accordance with the input electricity cost. They arrive at a model for power consumption of virtual machines hosted on a cloud infrastructure; this power consumption model is then used in calculating the cost of operation of the service. A cloud instance leased by a consumer is billed based on the cost of operation, and its resource utilization.
Conventional metering of cloud services is based on a fixed tariff rate (e.g., as defined in an SLA). The weakness of this model is that consumers are billed based on a predefined tariff rate, regardless of load and cost. “Smart metering” of services on the other hand would enable service providers to effectively manage computing resources with tariff plans accounting for the load condition on the infrastructure as well as the dynamic, time-varying nature of the input costs (especially electrical power).
Pay as you go (a/k/a static pricing) also does not suffice if service providers wish to provide bulk discounts or other economic incentives. It also is inadequate if there is a need to achieve congestion control, more uniform usage, or such objectives (as is done with time-varying highway tolls).
Power cost is often the largest component of an IT service’s lifecycle costs, and the major share of its operating expenditure (OpEx). In large IT systems, in addition to the power consumed by the computing devices themselves (all of which gets dissipated in the form of heat), the power consumed by the inevitable cooling systems is also significant. Google and others claim exceptionally low power usage effectiveness (PUE) values, but 2–3 seems common in the data center industry, meaning that in a typical data center, at least as much power is consumed in the cooling system as in the computing equipment itself.
Smart grids that price electrical power at time-varying rates, and also have associated demand-response programs that require electricity consumers to rapidly reduce power consumption on demand, are becoming common worldwide in the electrical power industry. It is thus likely that data center operators and vendors of cloud services will have to account for electrical power being available at time-varying rates, and further will have to devise mechanisms to reduce their power draws when asked. Smart grids are a hot topic of research these days, as is cloud computing. However, there is hardly any work to date which considers how the former would affect the latter. With smart grids becoming the new worldwide norm for electrical distribution, and power costs for IT systems being very high and also continually on the rise, it is likely that power-aware cloud services will become common in the foreseeable future.
By James Monroe