A new design for end-to-end proportional loss differentiation in IP networks Pablo J. Argibay-Losada * , Andrés Suárez-González, Cándido López-García, Manuel Fernández-Veiga Departamento de Ingeniería Telemática, Campus Universitario s/n, Vigo, Spain article info Article history: Received 19 February 2009 Received in revised form 20 November 2009 Accepted 2 December 2009 Available online 11 December 2009 Responsible Editor: A. Pitsillides Keywords: Quality of service Loss probability End-to-end performance Network architecture Proportional differentiation abstract This paper describes the algorithms and the architecture of a network able to provide end- to-end proportional packet loss probabilities at the flow level. We show that the combina- tion of a simple classification technique at the sources, and a network core having two internal service classes, is sufficient to achieve proportional service without the need to deploy coordinated, complex per-hop scheduling schemes or signaling protocols, which is the conventional approach. The proposed architecture is complementary to any differen- tiation algorithm used by the routers. Our results show that any network endowed with some internal service classes with respect to packet loss probabilities can be exploited to build a set of external service classes with end-to-end and per-flow guarantees. Ó 2009 Elsevier B.V. All rights reserved. 1. Introduction Packet forwarding in the Internet substantiates the best-effort service model, whereby routers do not keep state information for any of the active traffic flows and every packet receives the same common service. This key architectural principle is behind the unparalleled growth in size, bandwidth and data types carried by the network, but precludes better resource allocation for applications with quality of service (QoS) demands. The fundamental problem for providing multiple ser- vice classes in a packet network has always been the scala- bility of the architecture and of the routers’ algorithms. In the 1990s, IntServ and DiffServ [1] emerged from within the IETF as the two frameworks for building a network core with differentiated services. In IntServ, the applications could obtain even the strictest QoS requirements, since the architecture dictates per-flow, end-to-end resource reservations. As a consequence of this postulate, the routers must maintain and process data and control state for every flow of packets, participate in complex signaling procedures, and cooperate with routers in the same or in other domains in order to support end-to-end service guarantees. Therefore, IntServ cannot be deployed over the global, decentralized Internet, and was never seriously considered for adoption. DiffServ arose as a simpler, more scalable, manageable, and easily deployable solution for service differentiation in IP networks. Its premise is that individual flows with similar QoS requirements can be aggregated in larger traffic groups, called macroflows, that use a certain set of forwarding rules at the core routers, the per-hop-behavior (PHB). Thus, DiffServ can exploit the statistical multiplexing gains of aggregation, it does not re- quire any signaling protocol (i.e., there are no reserva- tions), and it is a distributed architecture, because the PHBs have strictly local semantics. The basic drawback is that the network cannot enforce a specific level of QoS to individual flows within an aggregate. Even the actual qual- ity of service received by a macroflow cannot be predicted accurately. 1389-1286/$ - see front matter Ó 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.comnet.2009.12.002 * Corresponding author. Tel.: +34 986813460; fax: +34 986 812116. E-mail address: pargibay@det.uvigo.es (P.J. Argibay-Losada). Computer Networks 54 (2010) 1389–1403 Contents lists available at ScienceDirect Computer Networks journal homepage: www.elsevier.com/locate/comnet