Kim, Peter; (1998) Resource Reservation in Shared and Switched Demand Priority Local Area Networks. Doctoral thesis (Ph.D), UCL (University College London).

Text Resource_reservation_in_shared.pdf Download (19MB)

Abstract

Packet switching data networks such as the Internet are migrating towards Integrated Services networks. To provide end-to-end service guarantees across those networks requires supporting mechanisms on all links along the data path including Local Area Networks (LAN) which are typically deployed at the leaves of the Internet. There is however no standard mechanism for building advanced services in existing LANs because the medium access mechanisms of these technologies differ. This dissertation is about providing Integrated Services in IEEE 802.12 networks. 802.12 is the standard for a shared 100 Mbit/s LAN. Its Medium Access Control (MAC) protocol is called Demand Priority. In this work, we have proved that the Guaranteed and the Controlled Load service proposed for a future multi-services Internet, can be provided across shared and switched 802.12 LANs, even when the network is overloaded with best effort traffic. This is achieved using resource reservation with admission control and based on the Integrated Services Packet Network (ISPN) framework. The key design constraints of our reservation scheme were the variable data throughput in 802.12 networks and the fact that hubs are not able to identify and isolate single data connections. We found that the Demand Priority signalling overhead may have a significant impact on the network performance when shared topologies become large or small sized data packets are used for data transmissions. To describe this overhead, a theoretical analysis is performed in which we derive results for topology and physical layer specific network parameters. Measurements in different test networks were used to confirm these results. The following part of the dissertation defines the admission control conditions for the Guaranteed service. When used with the parameters derived in the analysis, we find that these conditions enable us to accurately compute the minimum network throughput and thus the resource allocation limit. We also studied the delay characteristics and how network resource can be partitioned. The Controlled Load service was designed based on traffic aggregation and simple static priority scheduling within switches. This ensures low implementation costs and a deployment in existing or next generation LAN switches.

Download activity - last month

Export as JSONCSVXML

Download activity - last 12 months

Export as XMLCSVJSON

Downloads by country - last 12 months

Export as JSONCSVXML

Archive Staff Only

View Item