基于边缘云的动态和抢占式任务卸载调度算法研究

边缘云计算系统被广泛用于支持各种计算服务。针对边缘云计算环境中的任务卸载调度问题,考虑边缘云系统下的动态性和抢占式任务卸载调度，提出一个基于贪婪模拟退火启发式算法的在线卸载框架(SAOF),根据任务所需的传输延迟以及计算时间，进行周期性的卸载和调度计算，考虑独立任务的随机到达性和资源的异构性，动态地将新到达的任务分配到合适的目的地(边缘服务器或云服务器),并根据每个任务的延迟敏感性，抢占式地为其分配计算资源，使所有任务的总加权响应时间最小化。最后，在多组参数组合下生成测试实例并进行性能评估实验，将SAOF算法与3种优秀的卸载调度优化算法(Selfish算法、Nearest算法和OnDisc算法)进行对比，实验结果表明，SAOF算法能更有效降低所有任务的总加权响应时间。

Stochastic and Preemptive Task Offloading for Edge-cloud Computing

The edge-cloud computing systems are widely used to support various computation services. Aiming at the problem of task offloading and scheduling in edge-cloud computing environments, a stochastic task offloading problem in the edge-cloud computing system is considered. A greedy simulated annealing heuristic algorithm based online offloading framework(SAOF)is proposed for the problem under study, which periodically schedules and offloads tasks according to the required transmission delays and computation times. In consideration of the stochastic arrival of independent tasks and the heterogeneity of resources, SAOF assigns tasks to the appropriate destination (edge servers or cloud servers) dynamically and allocates computing resources to each task preemptively according to its latency-sensitivity. The goal is to minimize the sum of weighted response times of all the tasks. Testing instances are generated under the combination of various parameter settings for evaluation experiments. SAOF is compared with three excellent scheduling optimization algorithms(Selfish algorithm, Nearest algorithm and OnDisc algorithm). Experimental results indicate that SAOF can more effectively reduce the total weighted response time of all tasks.