CONTROL OF MULTIRATE SYNCHRONOUS STREAMS IN HYBRID TDM ACCESS NETWORKS

Authors
Citation
R. Bolla et F. Davoli, CONTROL OF MULTIRATE SYNCHRONOUS STREAMS IN HYBRID TDM ACCESS NETWORKS, IEEE/ACM transactions on networking, 5(2), 1997, pp. 291-304
Citations number
42
Categorie Soggetti
Engineering, Eletrical & Electronic","Computer Science Hardware & Architecture
ISSN journal
10636692
Volume
5
Issue
2
Year of publication
1997
Pages
291 - 304
Database
ISI
SICI code
1063-6692(1997)5:2<291:COMSSI>2.0.ZU;2-J
Abstract
An access multiplexer is considered, serving a TDM output channel with a hybrid structure, characterized by the presence of both isochronous (circuit-switched) and asynchronous (packet-switched) traffic, The is ochronous traffic is, in turn, subdivided into several classes, which are distinguished according to their speed, in order to model a multir ate, multitraffic environment, The aim of the paper is to define a con trol scheme for the allocation of the output link bandwidth, in order to realize two objectives: 1) minimize call blocking probability for t he isochronous traffic and packet loss probability, and 2) meet qualit y of service requirements for both traffic types as closely as possibl e, The proposed scheme is based on a hierarchical control structure, w here the description of the dynamics of the isochronous traffic is giv en in terms of continuous-time Markov chains and the packet rejection rate is approximated by means of the stationary distribution of an M/P areto/1/K queue by exploiting the large difference in time scales betw een the isochronous flows and the asynchronous one, There are two cont rol levels: 1) a ''fast'' one acting on the admission of isochronous c alls by means of distributed decisional agents (one for each traffic c lass) which operate with the time scale of the connection request proc ess dynamics, and 2) a slower one, playing the role of a coordinator i n the hierarchical scheme, which periodically recomputes a set of para meters (related to the bandwidth allocation) by numerically solving a parametric optimization problem, where real-time information as well a s traffic statistics are taken into account, Numerical results are pro vided, to investigate the performance of the control architecture.