Advanced Replacement Algorithms and In-Network Caching Optimizations for Information-Centric Network
| Autor: | Hong-Bin Tsai, 蔡宏彬 |
|---|---|
| Rok vydání: | 2018 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 106 As the next-generation Internet architecture, Information-Centric Network (ICN) has emerged to shift the current address-based Internet paradigm to a content-based one. In order to facilitate content dissemination in ICN, each router in the network is equipped with storage to implement in-network caching. With limited capacity to serve the enormous volume of contents transmitted in the Internet, a caching strategy is employed to manage contents in ICN routers. However, related research to improve performance of in-network caching solely focuses on the selection of a router or routers for content placement. The replacement policy of an individual router, however, is still Least-Recently-Used (LRU), a replacement algorithm that was proposed half a century ago, without a further discussion. In this dissertation, we carefully study replacement algorithms and propose our improvements with novelty. Moreover, we further extend our findings to collaborate with caching strategy in ICN and achieve significant improvements to the performance of in-network caching. Caching is widely employed in between tiers of computer memories to boost system performance. For example, cache in a processor stores data for the registers to reduce the latency of reading from slower memory; web cache stores data for clients to reduce the latency of reconstructing query results from database. By keeping popular data in fast and compact memory, costs from accessing large and slow storage are saved. A cache space, however, will be saturated with data after loads of operations. If a new data is fetched and to be inserted into cache under this circumstance, the cache has to evict an object and makes room for the coming data. The problem of victim selection that optimizes performance is hence defined as a replacement problem. The first part of this research digs into the paging subsystem of virtual memory, a special case of caching, and proposes a time-shifting technique that dramatically outperforms LRU. Contents in ICN are identified via names instead of addresses. Therefore, the request for a content does not necessarily be served by the content source. On receiving a request for a content, an ICN router checks if the content is in possession first. If there is a cache hit, the router forwards the content to the client directly. As such, contents may be fetched from routers nearby so that the network performance is improved. However, the most commonly acknowledged strategy for in-network caching, Leave-Copy-Everywhere (LCE), leaves replicas redundantly so that the cache utilization is inefficient. The second part of this research aims to improve the performance of ICN by considering caching strategy and replacement policy jointly. We propose a novel caching scheme for ICN that significantly reduces the average latency of content delivery over the network. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|