Delay-Sensitive Task Pre-Migration in SDN-Based Computing Power Networks

The prevalence of computational-intensive tasks stimulate the emergence of Computing Power Networks (CPN), where Software Defined Network (SDN) is utilized for flexible control. Considering the coexistence of in-execution tasks in the network and the arrival of emergency tasks, pre-migration is essential for reserving computing resources in advance, while it results in extra transmission cost in data plane and control cost in control plane. In this paper, we propose an analytical framework of task pre-migration policy in the SDN-based CPN to reduce system cost while meeting the delay budget. Specially, we consider the delay in SDN control plane induced by pre-migration flows, which involves <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\ell_{0}-\mathbf{norm}$</tex> and causes the problem non-convex and discontinuous. To overcome the challenges, we propose an iterative approximation algorithm to handle <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\ell_{0}-\mathbf{norm}$</tex> constraints, which transforms the original problem into a continuous function. Then we propose a dynamic programming based algorithm to minimize the approximation error, aiming at delay guarantee. The superiority of the proposed algorithm is verified through simulations, especially for heavy task offloading.