面向5G边缘计算的短视频缓存策略研究

Research on Short Video Caching Strategy for 5G Edge Computing

针对用户终端在移动过程中点播短视频而引起的响应时延问题,提出一种面向5G边缘计算的长短时记忆网络模拟退火算法缓存策略(Cache Strategy of Simulated Annealing Algorithm for Long and Short-term Memory Network,LSTM-SA)。该缓存策略采用“端-边-云”系统协同架构、网络模型、基于长短时记忆网络模拟退火算法等。首先,面对用户终端在不同MEC服务器之间移动时短视频请求刷新快的特点,基于“端-边-云”系统协同架构,以减少传输时延和优化用户体验质量(Quality of User Experience,QoE)为目标建立网络模型;其次,通过长短时记忆网络模型对用户终端移动路径进行预测,并寻求传输时延最小的局部最优解;最后,采用模拟退火算法对局部最小传输时延进行迭代更新,并以Metropolis准则为依据,在所有局部最优解空间中寻找全局最优解,从而保证将短视频内容缓存到合适位置。仿真结果表明:在“端-边-云”系统协同架构下,当用户终端在不同MEC服务器之间移动时,LSTM-SA缓存策略能够有效降低传输时延,提升用户体验质量,在短视频请求刷新快的场景能够实现顺畅连续播放且有效缓解网络拥塞问题。

Aiming at the response delay problem caused by the short video on demand by the user terminal during the movement,this paper proposes a 5G edge computing-oriented long and short-term memory network simulated annealing algorithm caching strategy(LSTM-SA).The“end-side-cloud”system collaborative architecture,network model,and simulated annealing algorithm based on long-short-term memory network used in this caching strategy are explored.Firstly,when the user terminal moves between different MEC servers,the short video request refreshes quickly,based on the“end-side-cloud”system collaborative architecture,with the goal of reducing transmission delay and optimizing the quality of user experience(QoE),establishing a network model;Then,it predicts the movement path of the user terminal through the long and short-term memory network model and seek the local optimal solution with the smallest transmission delay;Finally,by using simulated annealing algorithm to iteratively update the local minimum transmission delay,based on the Metropolis criterion,it searches for the global optimal solution in all the local optimal solution spaces,so as to ensure that the short video content is cached in a suitable location.Simulation results indicate that,under the“end-side-cloud”system collaborative architecture,when the user terminal moves between different MEC servers,the LSTM-SA caching strategy can effectively reduce the transmission delay and improve the quality of user experience.The LSTM-SA caching strategy can smoothly and continuously play in scenes where short video requests are refreshed quickly and effectively alleviate network congestion.