【摘要】 针对边缘计算带宽限制导致的实时流数据处理计算效率低下的问题,提出一种迭代优化算法FFS+IPFS,通过对应用负载的实时监控,实现合理的边缘节点任务部署,支持实时流数据处理任务。首先,利用贪心算法进行全局任务分配,通过贪心的算法得到一个近似最优的结果;然后,基于监控到的实时任务信息,通过迭代优化进行局部调优,使得同一数据流的任务可以被部署在相近的边缘节点,从而有效减少任务通信的开销。在不同场景下,平均时延相比其他主流算法可降低23%。大量的模拟实验结果表明,所提算法可以实现有效的资源调度,支持边缘计算场景下高效的实时流数据处理应用。更多还原

【Abstract】 Aiming at the low efficiency problem in real-time stream data processing at the edge due to bandwidth limitation,a two-step iterative optimization algorithm named FFS+IPFS(Find Feasible Solution+ ImProve Feasible Solution)was proposed to utilize the communication patterns between different tasks:a greedy approach was first used by FFS for global task assignment to achieve near optimal result;Iterative method was then employed by IPFS to optimize local results,so that tasks from the same application could be placed closely. In different scenarios,the latency can be reduced significantly by 23% compared with other SOTA(state-of-the-art)methods. Extensive experiment results show that FFS+IPFS can achieve very competitive results when applied in real-time data processing tasks in edge computing.更多还原