全称：IEEE Global Communications Conference
报告题目：Design and Evaluation of A Prediction-based Dynamic Edge Computing System
报告摘要：We investigate a mobile edge computing environment where edge computing nodes provide their computationcapacities to process the computation intensive tasks submittedby end users. We introduce aCloudlet Assisted Cooperative TaskAssignment (CACTA)system that organizes edge nodes that aregeographically close to a user into a cluster to collaborativelywork on the user’s tasks. The system enables a user to minimizehis/her total cost which is a weighted combination of latency(i.e., the task’s completion time), and the costs incurred inworking on the task. The total cost captures the tradeoff thatthe user would like to make between latency and computingrelated costs. It is challenging for the system to find an optimalstrategy that assigns workload to edge nodes to meet the user’soptimization goal, due to the time-varying available capacitiesand the mobility of edge nodes. To address the challenge, wemodel the system as a discrete time system in which each edgenode’s capacity and cost vary over different time slots, andthe system assigns parts of the task to the edge nodes in thecluster over time. We introduce a prediction-based dynamictask assignment algorithm, referred to PA-OPT, that assignsworkload to edge nodes in each time slot based on the predictionof their capacities/costs and an empirical optimal allocationstrategy which is learned from an offline optimal solution fromhistorical data. Then we apply our system design to a videodata analysis application, and conduct extensive simulationsdriven by a Google cloud data trace. We have demonstratedthat our proposed algorithm/system achieves significantly higherperformance than several other algorithms, and especially itsperformance is very close to that of an offline optimal solution.