Emergent Properties of the BT SDH Network
BT Technology Journal
Edge-to-edge measurement-based distributed network monitoring
Computer Networks: The International Journal of Computer and Telecommunications Networking
A decade of Internet research -- advances in models and practices
BT Technology Journal
Simulation with real world network stacks
WSC '05 Proceedings of the 37th conference on Winter simulation
Benchmarks for DDoS defense evaluation
MILCOM'06 Proceedings of the 2006 IEEE conference on Military communications
Hi-index | 0.00 |
While network simulations for congestion control stud-ieshave often varied traffic loads and protocol parameters,they have typically investigated only a few topologies. Themost common is by far the so-called "barbell" topology.In this paper we argue, first, that the barbell topology isnot representative of the Internet. In particular, we reportthat a measurable fraction of packets pass through multiplecongestion points. Second, we argue that the distinctionbetween the "barbell" topology and more complex topologiesis relevant by presenting a scenario with multiple congestionpoints that exhibits behavior that seems unexpectedbased on intuition derived from the barbell topology (inparticular, a TCP-only system that exhibits behavior technicallyconsidered "congestion collapse"). We make thelarger argument that the typical methodology currently acceptedfor evaluating network protocols is flawed. Finally,we briefly comment on some issues that arise in designinga simulation methodology that will be better suited to comparisonof network protocol performance.