Provision of Quality-of-Service (QoS) guarantees is an important and c
hallenging issue in the design of integrated-services packet networks.
Call admission control is an integral part of the challenge and is cl
osely related to other aspects of networks such as service models, sch
eduling disciplines, traffic characterization and QoS specification. I
n this paper we provide a theoretical framework within which call admi
ssion control schemes with multiple statistical QoS guarantees can be
constructed for the Generalized Processor Sharing (GPS) scheduling dis
cipline. Using this framework, we present several admission control sc
hemes for both session-based and class-based service models. The theor
etical framework is based on recent results in the statistical analysi
s of the GPS scheduling discipline and the theory of effective bandwid
ths. Both optimal schemes and suboptimal schemes requiring less comput
ational effort are studied under these service models. The QoS metric
considered is loss probability.