Abstract
Distributed real-time systems require guaranteed ographics. communication, where guarantees may be in terms of a maximum delay, minimum bandwidth and so on. In order to be able to provide these guarantees to applications, resources must be allocated along the path of connections. If resources are allocated for many connections, even though they may not be simultaneously sending data, over allocation of resources occurs. Resource sharing techniques are employed to conserve resources. Sender Sharing Groups, discussed in this thesis, is a mechanism for resource sharing, where resources along the path of connections from the sender are shared. Whenever connections are known to not carry data simultaneously, resources can be conserved by grouping such connections into Sender Sharing Groups. The sender agrees to send over not more than a predefined number of connections at any one time and the system guarantees the promised performance for the connections, as long as the sender does not violate this agreement. We discuss the mechanism of establishing and tearing down connections in a Sender Sharing Group. We also discuss the design of a two-level (independent and shared) politer, required for policing of sharing group data. The admission probability, policing time and admission time are measured in cases with and without Sender Sharing Groups.
Narayanan, Vidya (1998). Sender sharing groups in real-time communication. Master's thesis, Texas A&M University. Available electronically from
https : / /hdl .handle .net /1969 .1 /ETD -TAMU -1998 -THESIS -N36.