A Performance Model and Measurement Framework for DiffServ Implementations

Providers of data network services are in search of new solutions for service offerings over integrated service packet networks which can accommodate voice and video traffic in addition to data traffic. The goal is to provide users with cutting-edge services such as IP telephony, video conferencing, video on demand, telemedicine, etc. As the traditional design of data networks has had little or no support in providing the required traffic characteristics to voice and video streams, architectures for quality of service (QoS) have been defined. Of these architectures, differentiated services (DiffServ) is seen as a significant contributor to the migration of current data networks to next-generation networks capable of providing a broader range of services. However, the way they have been defined, these architectures lack precise formal specifications, and therefore, their performance is strongly related to the quality of each implementation. This paper presents an abstract model for DiffServ with assured forwarding (AF) per-hop behavior (PHB). The model is meant for a better illustration of the dynamics of QoS parameters. After a thorough review of existing methodologies for performance measurements for DiffServ, this paper proposes a dynamic model for AF and a framework for its performance measurement, which is illustrated by a number of experiments run on a live network (NCIT*net 2). This paper is the starting point for the detailed specification of DiffServ PHBs. Experimental results illustrate, verify, and justify the model introduced.