On characterizing BGP routing table growth

  • Authors:
  • Tian Bu;Lixin Gao;Don Towsley

  • Affiliations:
  • Bell Laboratories, Lucent Technology, Holmdel, NJ;Department of Electrical and Computer Engineering, University of Massachusetts, Amherst, MA;Department of Computer Science, University of Massachusetts, Amherst, MA

  • Venue:
  • Computer Networks: The International Journal of Computer and Telecommunications Networking - Special issue on The global Internet
  • Year:
  • 2004

Quantified Score

Hi-index 0.00

Visualization

Abstract

The sizes of the BGP routing tables have increased by an order of magnitude over the last six years. This dramatic growth of the routing table can decrease the packet forwarding speed and demand more router memory space. In this paper, we explore the extent that various factors contribute to the routing table size and characterize the growth of each contribution. We begin with measurement study using routing tables of Oregon Route Views server to determine the contributions of multi-homing, load balancing, address fragmentation, and failure to aggregate to routing table size. We find that the contribution of address fragmentation is the greatest and is three times that of multi-homing or load balancing. The contribution of failure to aggregate is the least. Although multi-homing and load balancing contribute less to routing table size than address fragmentation does, we observe that the contribution of multi-homing and that of load balancing grow faster than the routing table does and that the load balancing has surpassed multihoming becoming the fastest growing contributor. Moreover, we find that both load balancing and multi-homing contribute to routing table growth by introducing more prefixes of length greater than 17 but less than 25, which is the fastest growing class of prefixes. Next, we compare the growth of the routing table to the expanding of IP addresses that can be routed and conclude that the growth of routable IP addresses is much slower than that of routing table size. Last, we demonstrate that our findings based on the view derived from the Oregon server are accurate through evaluation using additional 15 routing tables collected from different locations in the Internet.