Abstract:
Multimedia transmission requires timely and ordered delivery of data with possibly
no retransmission. As the number of users of multimedia applications is increasing
day by day, overloaded servers and network congestion are also growing rapidly.
These are among the main reasons for delay in accessing content from a server via a
network path. Multimedia applications require a particular Quality of Service (QoS)
level to be maIntained to fulfill users' satisfaction. But it has another goal:
maximizing the earned revenue from the users for the network owner. As resources
are not unbounded, some selection criterion is essential when the number of users
requesting resources for multimedia data transmission is large and their requested
QoS is to be maintained. A particular QoS can be maintained only when necessary
server and network resources is reserved. Current technology like best-effort service
is unsuitable for above-mentioned requirements because all packets compete equally
for network resources. The necessity to develop better QoS solutions to address these
issues of multimedia transmission has led Internet Engineering Task Force (IETF).
In this thesis we designed an admission controller for an Enterprise Network which
maximizes the earned revenue from the admitted users among many of them
requesting multimedia data transmission with guaranteed QoS in terms of both server
and network resources. Our admission controller considers the selection of both a
server from several alternatives as a source of multimedia data and a delivery route
along with reservation of resources in the selected server and on the delivery route so
that the satisfaction of all admitted users is achieved. We redefine the Service Level
Agreement (SLA), a contractual agreement between a user and the network owner
and specifying several levels of proposal in terms of QoS. Protocols for data
transmission in different stages of the admission controller have also been pointed
out.
We designed necessary admission control algorithm and analyzed its complexity.
Finally we simulated the admission controller, and performance data are collected
extensively from a Java simulation program. We have also analyzed the performance
11
data with proper reasoning. The validation of the experiment has been presented by a
graph theoretical analysis of the network and servers used for the simulation.
