RFC 2544 performance evaluation and internal measurements for a Linux based open router

Abstract

Recent technological advances give a good chance to do something really effective in the field of open Internet equipments, also called Open Routers (ORs). Some initiatives have been activated since the last few years to investigate the OR and related issues. But despite these activities, large interesting areas still require a deeper investigation. This work tries to give a contribution by reporting the results of an in-depth activity of optimization and testing realized on a PC Open Router architecture based on Linux software and COTS hardware. The main target approached in this paper has been the forwarding performance evaluation of different OR Linux-based SW architectures. This analysis has been performed with both external (throughput and latencies) and internal (profiling) measurements. In particular, for what concerns the external measurements, a set of RFC2544 compliant tests has been proposed and analyzed.

Compressing IP forwarding tables

“…Crucially, small storage can only be attained if the leaf-push barrier is chosen according to (2). Strikingly, we found that precisely this setting is the key to fast FIB updates as well 2 .…”

“…If p < λ, then updating a single entry can be done in O(W ) time. If, on the other hand, p ≥ λ, then update visits at most W + 2 W −λ ≤ W + W H 0 nodes, using that λ ≥ W − lg( W H 0 ) whenever λ is as (2). In summary, under mild assumptions on the label distribution a prefix DAG realizes the Shannon-entropy up to a 2 Note that (2) transforms into (1) at maximum entropy.…”

Compressing IP Forwarding Tables: Towards Entropy Bounds and Beyond

“…Correspondingly, FIBs continue to expand both in size and management burden. As a quick reality check, the Linux kernel's fib_trie data structure [41], when filled with this many prefixes, occupies tens of Mbytes of memory, takes several minutes to download to the forwarding plane, and is still heavily debated to scale to multigigabit speeds [2]. Commercial routers suffer similar troubles, aggravated by the fact that line card hardware is more difficult to upgrade than software routers.…”

Compressing IP forwarding tables for fun and profit

“…Correspondingly, FIBs tend to grow large both in size and management burden, presenting a crucial bottleneck in the data-plane performance of routers and forcing operators into rapid upgrade cycles [2]. As a quick reality check, the Linux kernel's fib_trie data structure [3], when filled with 400,000 prefixes, occupies more than 20 Mbytes of memory, takes several minutes to download to the forwarding plane, and is still heavily debated to scale to multigigabit speeds [4]. Commercial routers suffer similar troubles, aggravated by the fact that line card memory is much more difficult to upgrade than software routers.…”