Latency equalization: a programmable routing service primitive

  • Authors:
  • Minlan Yu;Marina Thottan;Li Li

  • Affiliations:
  • Princeton University, Princeton, NJ, USA;Alcatel-Lucent, Murray Hill, NJ, USA;Alcatel-Lucent, Murray Hill, NJ, USA

  • Venue:
  • Proceedings of the ACM workshop on Programmable routers for extensible services of tomorrow
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

Today the Internet is the primary medium for deploying new real time services such as gaming and distributed online live music concerts. Different network services have different expectations from the routing infrastructure. Some network services require conventional routing paths optimized for low latency or low congestion. However, real-time interactive services such as online gaming and distributed live music performance require more than just low latency. They require Latency EQualization (LEQ) among participating users. Although LEQ could be performed by the client or the server, end-system techniques for estimating network conditions are often inaccurate. Instead, we argue that the network should provide a LEQ service. We propose a LEQ routing architecture that can leverage programmable hub nodes. By deploying a few flexible, well-placed programmable nodes to redirect traffic, we can flexibly support both latency equalized and low latency routing services simultaneously. For LEQ routing, programmable hub nodes provide services such as application packet identification, application level packet processing and latency equalized routing paths. Extensive simulation studies on provider network topologies show that using just a few programmable nodes we can achieve an 80% improvement in LEQ over the conventional architecture that uses shortest path routing.