The rapid growth of routing tables represents a major challenge facing the scalability of BGP and indeed the whole Internet infrastructure. In this paper, we introduce a novel distributed algorithmic scheme for partitioning the BGP routing table on multiple controller cards, where we exploit parallelism to enhance both the lookup speed and the scalability of the RIB (Routing Information Base). The proposed scheme increases the lookup performance by letting unrelated tasks, such as the Best Match Prefix (BMP) lookup and the BGP decision process to be executed in parallel at different controller cards. Simulations show that our proposal outperforms classical central lookup mechanisms with a reasonably acceptable cost, while it increases considerably the space scalability of the BGP routing table.
展开▼
