Decoupling different time scales of network QoS systems

  • Authors:

  • Jens Schmitt;Oliver Heckmann;Martin Karsten;Ralf Steinmetz

  • Affiliations:

  • Department of Electrical Engineering and Information Technology, Darmstadt University of Technology, Industrial Process and System Communication (KOM), Merckstr. 25, 64283 Darmstadt, Germany;Department of Electrical Engineering and Information Technology, Darmstadt University of Technology, Industrial Process and System Communication (KOM), Merckstr. 25, 64283 Darmstadt, Germany;Department of Electrical Engineering and Information Technology, Darmstadt University of Technology, Industrial Process and System Communication (KOM), Merckstr. 25, 64283 Darmstadt, Germany;Department of Electrical Engineering and Information Technology, Darmstadt University of Technology, Industrial Process and System Communication (KOM), Merckstr. 25, 64283 Darmstadt, Germany

  • Venue:
  • Computer Communications
  • Year:
  • 2002

Quantified Score

Hi-index 0.24

Visualization

Abstract

Providing quality of service (QoS) in large-scale networks like the Internet inherently needs to deal with heterogeneous network QoS systems. Therefore, the interworking between different network QoS systems is of high importance. In this paper, the interworking with respect to a basic characteristic of network QoS systems, the time scale of the system, is under investigation. The time scale of a network QoS system is its speed of reaction to individual requests for differentiated treatment of units of service. A slow time scale system will prefer requests to arrive with a low frequency and persist unaltered for a substantial period of time while a fast one is able to support much higher arrival rates of requests and is thus more amenable for short-lived units of service. Obviously, when overlaying a slow time scale QoS system over a faster one, there is no problem. However, and that is a more likely case, for the overlay of a fast time scale system on a slow one, there is a mismatch to be mediated at the edge between the two. The technique that is applied at an edge device for this mediation is called decoupling of time scales. Decoupling can also be viewed as aggregation of requests in time in contrast to spatial aggregation on the data path. In the paper, we develop an adaptive heuristic scheme to deal with decoupling and evaluate this scheme by extensive simulations.