Show simple item record Wang, Dabin en_US 2007-05-18T12:16:01Z 2007-05-18T12:16:01Z 1998-12-01T00:00:00Z en_US
dc.description.abstract In this thesis we cover two main topics, reliable network transmission protocol modeling and design. The first is a reliable transmission protocol modeling by the place stochastic Petri net. The second is retransmission timer design for TCP. The first consideration leads to a pursuit of the tightest bounds of the timing constraints given in a specification using functional optimization. For TCP, this leads to a requirement for traffic measurements and observations that resulted in two main results. First, we found the packet round trip time over IP has the self-similar property. To our knowledge, no one has found and considered this property about the round-trip time of transmitted packets over IP. Second, we attempted to estimate the parameters in Jacobson's dynamic retransmission timeout algorithm by making use of self-similarity parameter. In the recent years, many researchers reported that actual network traffic is self-similar in nature. However, effectively designing protocols that take self-similarity into account remains largely an open issue. There are two major problems: (1) What is the physical "explanation" for observed self-similar nature of measured traffic from today's packet networks? (2) What is the impact of self-similarity on network and protocol design and performance analysis? Walter Willinger, Murad S. Taqqu, Robert Sherman, and Daniel V. Wilson and related studies answered question (1). The study here is related to question (2). (Abstract shortened by UMI.) en_US
dc.format.extent 3228034 bytes
dc.format.extent 184 bytes
dc.format.mimetype application/pdf
dc.format.mimetype text/plain
dc.language en en_US
dc.language.iso en_US
dc.title Reliable network transmission protocol modeling and design en_US Electrical and Computer Engineering en_US Master of Science (M.Sc.) en_US

Files in this item

This item appears in the following Collection(s)

Show simple item record

View Statistics