Coupling caching and forwarding: benefits, analysis, and implementation
Coupling caching and forwarding: benefits, analysis, and implementation
复制标题
DOI：
10.1145/2660129.2660153
复制
发表时间：
2014-09
期刊：
影响因子：
--
通讯作者：
Giuseppe Rossini;D. Rossi
中科院分区：
文献类型：
--
作者：
Giuseppe Rossini;D. Rossi
最近的一场辩论围绕着普遍缓存的有用性展开，即可能为未来互联网的每一台路由器添加缓存功能。最近的研究反对它，理由是与当前的CDN场景相比，它提供的收益非常有限，因为在CDN场景中，缓存仅发生在网络边缘。在本文中，我们相反地表明，只有当元缓存(即，是否缓存传入对象)和转发(即，在缓存未命中的情况下将请求定向到哪里)决策紧密耦合时，泛在缓存的优势才会显现。总结我们的贡献，我们(I)表明，只要联合使用理想最近副本路由(INRR)转发和Leave a Copy Down(LCD)元缓存，就可以获得收益，(Ii)对iNRR转发策略进行建模，(Iii)提供两种任意接近iNRR行为的替代实现，以及(Iv)通过使我们的代码对社区可用来促进交叉比较。
A recent debate revolves around the usefulness of pervasive caching, i.e., adding caching capabilities to possibly every router of the future Internet. Recent research argues against it, on the ground that it provides only very limited gain with respect to the current CDN scenario, where caching only happens at the network edge. In this paper, we instead show that advantages of ubiquitous caching appear only when meta-caching (i.e., whether or not cache the incoming object) and forwarding (i.e., where to direct requests in case of cache miss) decisions are tightly coupled. Summarizing our contributions, we (i) show that gains can be obtained provided that ideal Nearest Replica Routing (iNRR) forwarding and Leave a Copy Down (LCD) meta-caching are jointly in use, (ii) model the iNRR forwarding policy, (iii) provide two alternative implementations that arbitrarily closely approximate iNRR behavior, and (iv) promote cross-comparison by making our code available to the community.