Cache replacement schemes for data-driven label switching networks

Abstract

The traffic on the Internet has been growing exponentially for some time. This growth is beginning to put stress on the current-day routers. On the other hand, switching technology, such as ATM, offers much higher performance than IP routing, but it is not the global internetwork protocol. As a result, the label-switching network which combines IP routing with switching technology has emerged. In the data-driven label switching, flow classification and cache table management are needed. Flow classification classifies packets into switching and nonswitching ones and cache table management maintains the cache table that contains information for flow classification and label switching. More specifically, cache table management affects the performance of the label-switching network considerably as well as the flow classification. This is because the bigger cache table makes it possible for more traffic to be sent by switching and maintains lower setup cost. However, the size of the cache table is restricted by router resources. Therefore, for efficient cache table management, studies on the cache replacement scheme are needed. This scheme must consider the characteristics of the real Internet traffic generated by the user. We evaluate basic cache replacement schemes: first in first out (FIFO), least flow count (LRC) and least recently used (LRU). Then we propose and evaluate weighted priority LRU and L2A schemes. The proposed schemes exhibit a better performance than those basic schemes.