您当前访问的的浏览器版本过低,为了给您带来更好的体验,建议您升级至Edge浏览器或者推荐使用Google浏览器
取消

基于多维梯度的卫星集群高可靠协同计算方法

作者:马步云, 任智源, 李赞 阅读量:423

基于多维梯度的卫星集群高可靠协同计算方法

马步云, 任智源, 李赞
(西安电子科技大学综合业务网理论及关键技术国家重点实验室,中国 西安 710071)

摘要:在分布式低轨(LEO)卫星集群中进行端到端业务计算是当前的研究热点。但是,传统分布式LEO端到端业务计算采用的集中式调度方法,存在任务执行成功率难保障、中心节点难选取等问题。针对上述问题,提出了一种新型分布式端到端业务计算技术,研究了基于时空扩展图模型的LEO拓扑动态性屏蔽方法,并提出了基于可靠性梯度选择的端到端业务计算理论模型与分布式业务调度算法。该算法可通过综合考虑LEO节点的多维信息来选取参与任务处理的路径节点,并逐跳完成业务的计算与传输。仿真结果表明,与集中式业务调度相比,分布式业务调度以较小的时延增加为代价,有效提升了任务的执行成功率。 
关键词:低轨卫星;分布式;端到端;协同计算


Multi-Dimensional Gradient Based High-Reliability Collaborative Computing Method for Satellite Network


MA Buyun, REN Zhiyuan, LI Zan
(State Key Laboratory of Integrated Services Networks, Xidian University Xi’an 710071, China)

Abstract: It has become a hot research field to study end-to-end task computing in low earth orbit (LEO) network. However, there exist difficulties to guarantee the task execution success rate and elect the management node caused by the centralized task scheduling. To solve the above problems, the distributed end-to-end information processing technology is proposed. At first, the dynamic of the LEO network is shielded based on the time expanded graph model; then, the end-to-end task computing model is constructed and the distributed task scheduling algorithm is proposed, which jointly considers the multi-dimensional of nodes to select the processing nodes. The simulation results validate that the distributed task scheduling effectively improves the task execution success rate at the cost of little latency performance.
Keywords: low earth orbit; distributed; end-to-end; collaborative computing

在线PDF浏览:PDF