由于边缘云没有比中心云更强大的计算处理能力,在应对动态负载时很容易导致无意义的扩展抖动或资源处理能力不足的问题,所以在一个真实的边缘云环境中对微服务应用程序使用两个合成和两个实际工作负载进行实验评估,并提出了一种基于负载预测的混合自动扩展方法(predictively horizontal and vertical pod autoscaling,Pre-HVPA)。该方法首先采用机器学习对负载数据特征进行预测,并获得最终负载预测结果。然后利用预测负载进行水平和垂直的混合自动扩展。仿真结果表明,基于该方法所进行自动扩展可以减少扩展抖动和容器使用数量,所以适用于边缘云环境中的微服务应用。
针对低轨卫星单星处理任务成本高,卫星网络节点近期故障次数不同的问题,提出了基于强化学习的低轨卫星星间计算卸载与资源分配方法。首先,通过对卫星网络节点进行筛选,将符合需求的可用卫星加入边缘卫星群组。并结合任务信息与边缘卫星节点信息来建立本地计算模型与卸载计算模型。然后,以最小化时延与能耗,最大化可信值为优化目标建立优化问题。最后,利用DDPG算法对卸载决策与计算资源分配联合求解。仿真表明,所设计的星间计算卸载策略相比于单星处理任务,能减少51.46%任务时延,与DQN算法相比,能减少26.11%时延。A reinforcement learning-based method for inter-satellite computation offloading and resource allocation of low-orbit satellites is proposed to address the high cost of single satellite processing tasks and the varying frequency of recent failures in satellite network nodes. Firstly, by filtering the satellite network nodes, available satellites that meet the requirements are added to the edge satellite group. And combine task information with edge satellite node information to establish local computing models and offloading computing models. Then, an optimization problem is established with the objective of minimizing latency and energy consumption, and maximizing the credibility value. Finally, the DDPG algorithm is used to jointly solve the unloading decision and computing resource allocation. The simulation shows, compared to single-star processing tasks, the designed inter-satellite computing offloading strategy can significantly reduce task costs and increase the reliability of offloading.
暂无评论