面向移动端基于社交关系的图片缓存替换算法

随着移动终端深入人们的生活,移动社交APP得到了广泛使用。在移动社交APP中往往会使用大量的图片资源,如微信朋友圈、Instagram的图片分享等。在APP中浏览图片会消耗较多的网络流量,影响加载速度,因此大部分APP采用首先显示缩略图,根据用户需求再加载原图的策略。在服务器端也采用缓存技术来加快缩略图产生时间,减少磁盘I/O。但是,当前的缓存机制更多关注的是缓存的访问频率、最近访问时间等因素,并没有过多关注数据生成用户之间的社交关系,也没有考虑移动用户对缩略图和原图的不同访问模式。把缓存划分为两个部分:缩略图缓存区和原图缓存区,提出了基于社交关系的图片缓存替换算法,在传统缓存替换算法的基础上增加用户的社交关系以及缩略图和原图的关联关系,通过计算图片的缓存价值进行缓存替换。实验表明,所提出的基于社交关系的图片缓存替换算法对于缩略图和原图的缓存命中率都有明显提高。

Image cache replacement algorithms based on social relationships for mobile terminals

As mobile terminals have great involvement in people’s lives, the mobile social APPs have been widely used. A large number of image resources are used in mobile social APPs, such as WeChat’s circles and Instagram’s picture sharing. While browsing pictures in APPs, we can consume more network traffic and the loading speed is affected. So most APPs display thumbnails first, and then load the original images according to the needs of users. On the server,we usually use caching techniques to speed up the thumbnail generation and reduce the I/O of the disk. However, current cache mechanism concerns more about the cache access frequency, last access time and other factors, ignoring social relationships between users and the different access modes how mobile users access the thumbnails and the original pictures. So we divide the cache buffer into two layers: the original cache buffer and the thumbnail cache buffer, and propose two image cache replacement algorithms based on social relationships, which add social relationships among users and relationships between the thumbnails and the original pictures to the traditional cache replacement algorithm, and replace the cache objects by computing the cache value.Experimental results show that the proposed image cache replacement algorithms based on social relationshipscan improve cache hit ratio of the thumbnails and the original images.