Summary: | 博士 === 中興大學 === 資訊科學與工程學系所 === 99 === Performance modeling plays an important role in designing and analyzing communication techniques.
When communication techniques promise to guarantee QoS for the running applicaations, accurate
performance prediction might be helpful during the deployment.
Recently, most emergeing frame-based communication techniques restrict events to only occur at
definite time point through the predefined frame structure.
However, the widely adopting continuous-time queueing models, in which events can occur at any
time, are not capable of capturing the behavior of frame-based communication systems.
The aim of this dissertation is to evaluate frame-based communication systems using discrete-time queueing models
instead of continuous-time queueing models for better reflecting the frame characteistic.
Since either real-time traffic in WiMAX (Worldwide Interoperability for Microwave Access) network or safety traffic in
WAVE (Wireless Access in the Vehicular Environment) netwok has rigid demand in transmission,
accurate performance analysis is indeed beneficial for constructing these network environments.
And different types of discrete-time models are explored to investigate the effect of parameters on the WiMAX and WAVE
networks.
In the WiMAX network, Unsolicated Grant Service (UGS) and real-time Polling Service (rtPS), which are dedicated for
real-time applications, are discussed throughout.
The discrtete GI-D-c model having deterministic service time is first adopted for modeling UGS connections.
Second, the polling interval of unicast polling is proved to be geometric distributed, and the discrete GI-Geo-1 model
having geometric service time is utilized for rtPS connections.
In addition to polling interval, other required time intervals of rtPS connections then are studied in the
general discrete GI-G-1 model.
In the WAVE network, the safety messages transmission in the multihop vehicle-to-vehicle network is investigated.
The multihop transmission is modeled as a queueing network, which can be solved by Discrete-Time Markov Chain (DTMC).
In order to resolve the DTMC model efficiently, an approximation approach based on the decomposition approach is
presented.
|