当前位置: X-MOL 学术J. Parallel Distrib. Comput. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Deep learning inspired routing in ICN using Monte Carlo Tree Search algorithm
Journal of Parallel and Distributed Computing ( IF 3.4 ) Pub Date : 2021-01-06 , DOI: 10.1016/j.jpdc.2020.12.014
Nitul Dutta , Shobhit K. Patel , Vadim Samusenkov , Vigneswaran D.

Information Centric Networking (ICN) provides caching strategies to improve network performance based on consumer demands from the intermediate routers. It reduces the load on content server, network traffic, and improves end-to-end delay. The content requesters use an Interest packet containing the name of data to express their needs. If such Interest packets are routed efficiently, the end to end delay and throughput of the network could be improved further. This paper describes an efficient method of forwarding Interest packets to retrieve the requested content at the earliest possible time. Here the data source is found and considered as a single player game with content requester as its start state and location of the desired content as final or goal state. The Monte Carlo Tree Search (MCTS) algorithm is used for constructing the path from content requester to concerned data source. For performance evaluation, the proposed scheme is integrated with Leave Copy Down (LCD) and Leave Copy Everywhere (LCE), Cache Less for More (CL4M), and Probability based caching (ProbCache) In ns-3 simulation environment (ndnSim), all these are evaluated in terms of content search latency, server hit ratio, network load, overhead and throughput. Simulation observation reveals that the integration of MCTS significantly improves performance in regard to experimental parameters.



中文翻译:

使用蒙特卡洛树搜索算法的ICN中的深度学习启发式路由

信息中心网络(ICN)提供缓存策略,以根据中间路由器的消费者需求提高网络性能。它减少了内容服务器的负载,网络流量,并改善了端到端延迟。内容请求者使用包含数据名称的兴趣包来表达他们的需求。如果有效地路由此类兴趣包,则可以进一步提高网络的端到端延迟和吞吐量。本文介绍了一种转发兴趣数据包的有效方法,以尽早检索请求的内容。在这里,找到数据源并将其视为具有内容请求者的开始状态的单个玩家游戏,并将所需内容的位置作为最终或目标状态。蒙特卡罗树搜索(MCTS)算法用于构建从内容请求者到相关数据源的路径。为了进行性能评估,建议的方案与“留空复制”(LCD)和“无处不在复制”(LCE),“少用多取”(CL4M)和基于概率的高速缓存(ProbCache)集成在一起。在ns -3仿真环境(ndnSim)中,将根据内容搜索延迟,服务器命中率,网络负载,开销和吞吐量来评估所有这些。仿真观察表明,MCTS的集成显着提高了实验参数的性能。

更新日期:2021-01-18
down
wechat
bug