Energy-efficient dispatching policy for virus scanning as a service under N-version protection

摘要

We investigate energy-efficient dispatching policy for virus scanning as service (VSaaS) under N-version protection. Under N-version protection, one would dispatch a file-scan request to multiple service engines, in order to reduce the missed detection error cost. However, this also increases the energy consumption of service engines and the queueing delay. To manage this tradeoff, we aim to reduce the energy consumption of service engines and the missed detection error cost, while achieving low delay for VSaaS. We first define a penalty function as the combination of the energy consumption and the missed detection error cost. We then formulate the file-scan dispatching policy as a constrained optimization problem under the framework of Lyapunov optimization. Using the Lyapunov-drift-penalty function, we propose an online algorithm, which achieves the penalty function arbitrarily close to the minimum by increasing the control variable but at the cost of increasing the queue length. Simulation results indicate that the proposed algorithm is flexible to provide N-version protection. Using the algorithm, the cloud operator can dynamically tune the control variable in order to reduce the energy consumption and the missed detection error cost, while maintaining the queue stability.