A Markovian Queueing Model for End-to-End Delay Analysis in Computation Offloading System

Smart terminals can reduce processing delay by allocating computation-intensive tasks to the mobile edge and local servers in a computation offloading system. However, unreasonable computation offloading may increase processing delay due to the limited computation capacity. To solve this problem, this letter proposes a Markovian queueing model for the computation offloading system to analyze the end-to-end delay. We first model the offloading system as a queueing network and represent it by a high-dimensional Markov chain with the system state as the queue lengths. Meanwhile, we propose an algorithm to generate the state transition matrix and derive the steady probability. By exploiting the steady probability, we analytically express the end-to-end delay based on Little’s Law. Then, we formulate a one-dimensional search problem to minimize the average delay. Specifically, a ternary search algorithm is presented to effectively find the minimized end-to-end delay and achieve the balance of computation workloads between the mobile edge and local servers. Finally, numerical results verify the theoretical derivation and further demonstrate the proposed computation offloading strategy.