6. Conclusions and Future Work
In this paper, we devise the energy cost optimization architecture for IDC operator to minimize the energy cost under the job risk probability constraint. The jobs may be delay tolerant big data applications or data intensive MapReduce applications that demands large-scale infrastructures such as Internet data center to provide computing resources. Due to high time complexity of optimal security levels mapping scheme, a heuristic algorithm with polynomial time complexity is developed to select security levels for tasks. Then, we formulate the energy stochastic optimization problem and propose our ECM algorithm to schedule workloads taking the temporal diversity of electricity price into account. The ECM algorithm, which is based on Lyapunov optimization framework, offers provable energy cost and delay guarantees. It aggressively and adaptively seizes the timing of low electricity price to process tasks, and defers delay-tolerant tasks execution when the price is high.
Four reference algorithms are conducted in our experiments in comparison with our ECM algorithm in terms of energy cost and queuing delay. The experiments confirm the [O(1/V),O(V)] energy-delay tradeoff of ECM algorithm. However, the performance of ECM algorithm is close to the enumeration algorithm, but with lower time complexity. In a word, Extensive evaluation experiments based on the real-life electricity price demonstrate the effectiveness of our ECM algorithm.
As a future work, we plan to incorporate the big data scientific workflow scheduling method and delay guarantee into our energy cost optimization problem. Moreover, we are going to consider some new aspects in better usage of power in IDC, such as renewable energy, energy storage, battery and so on.
