Caching System Based On Next Requested Time Of Video Block

The video caching system is a key component of the online streaming system.The cache servers are usually set up at the edge of the network to relieve bandwidth pressure and improve the users'viewing experience by delivering video data to users directly.Caching algorithms are key factors to determine cache service performance,for the reason that they measure the importance of the different video blocks and store the most important data.Due to theirs simplicity and effectiveness,the LFU and LRU caching algorithms are still widely used in actual VoD system.They all simply take the users'historical data accesses as future data requests,and use the number of requests that have occurred(LFU)and the most recent request time(LRU)that have occurred to measure the importance of a video block,respectively.However,for streaming media service such as VoD,online users usually do not view the video blocks they have seen.According to the existing algorithm,the numbers of historical access or the time of the video blocks requested cannot truly and accurately reflect the future importance of the data blocks.Specifically,LFU can only seize the cache opportunities on a large time scale due to it caches video blocks according to their popularities;The LRU attempts to cache data on a small time scale according to the last request time,which couldn't achieve the desired results.In fact,online users at different viewing schedules typically view subsequent video blocks with a high probability.This unique perspective opens up new possibilities for optimizing cache performance.We can combine the viewing needs of both online users and offline users(on the road)to accurately assess the importance of data blocks and improving cache performance.However,due to the random departure/arrival of online/offline users,designing a more efficient caching algorithm by describing and combining the viewing requirements of online/offline users is an important theoretical and practical question and a challenging question.In this paper,we propose a cache-scheduling algorithm based on video block requested time.The main idea is to measure the importance of the video data block by estimating the NRT(Next Requested Time)of a video data block according to the actual subsequent viewing requirements of the online users and the statistical viewing demands of the offline users.In addition,our cache algorithm caches the video blocks with the most recent request time according to the optimal cache algorithm idea.The simulation based on actual business data shows that the proposed algorithm in this paper can capture the available cache opportunities on both large and small time scales,which greatly improves the efficiency of cache system.Specifically,the contributions of this article are as follows.(1)A calculation model of the next requested time of the video blocks is proposed to accurately evaluate the importance of the video blocks.The model simultaneously depicts and combines the viewing demands of both online and offline users,and can accurately measure the importance of video blocks at a finer granularity,laying the foundation for efficient caching.(2)A caching algorithm based on the next requested time of the video blocks is designed to achieve efficient cache scheduling.The algorithm calculates the next requested time of the video blocks in real time,and caches the video block with the latest requested time.(3)A large-scale simulation based on actual business data is completed.The numerical results show that the algorithm proposed in this paper can significantly improve the cache hit ratio,which is 34.8%and 212%higher by the LFU and LRU cache hit ratio.