Research On Modeling And Analyzing Of Networks End-to-End Communication Performance Based On QBD

End-to-end (e2e) communication is the problem of sending a sequence of messages from a sender to a receiver, through an unreliable communication network. It is the basis of adaptive communication protocols and distributed computing systems. Aiming at achieving some given performance goals, e2e adaptive communication protocols must be capable of inferring future packet losses and self-adjust their behaviors, in order to cope with the variability in the network conditions. These control mechanisms often rely on packet loss models that need to be accurate enough for analysis. Unfortunately, the behaviors of network communication are complex and unpredictable, due to the capacity differentiation of different communication links. The performance analysis of these systems, therefore, is enormously difficult, primarily because of complex and unpredictable network interactions that complicate the queueing mechanisms. Moreover, the mutual interference among e2e communication flows in wireless networks complicates their analysis. As a result, a new, systematic method for analyzing the performance of end-to-end communication system plays a key role in network optimization and management.Considering these problems in the unreliable communication network, this thesis proposes systematic mechanisms, and anaytical models for PLCN (Probabilistic Lossy Communication Networks), WLCS (wireless lossy channel system), E2E-CSLC (e2e channel system with lossy communication), and (x,S)-bottleneck cell in larg-scale wirlss networks. The main works include:1. The performance analytical model for e2e communication channel over probabilistic lossy communication networks (PLCN). The analytical mdel for packet loss processes must be capable of accurately describing and inferring the behaviors of packet loss. However, existing analytical models of packet loss cannot provide the dynamic relationship between states of lossy channel systems due to the limitation of state space. In this dissertation, an analysis on balance state of e2e communication channel over PLCN is performed firstly. We, then, discuss the process of modeling e2e communication channel over PLCN using QBD in detail. Based on our analytical model, some characteristics of e2e communication channel, such as the stability, and the packe-loss performance of channel, are analyzed. The results derived from the model are helpful for the further improvement of adaptive end-to-end communication protocols in the future.2. Probabilistic analysis for state reachability of wireless lossy channel systems (WLCS). We extend PLCN towards WLCS by considering the method of packet loss and the transition rule, and provide a technique to analyze the state reachabilities of WLCS using QBD. To study on state reachabilities of WLCS based on QBD, we prove that the state set,S0, is an attractor of WLCS model. Based on the analytical model, we can perform some quantitive analysis on the probabilities of the steady-state reachability and the transient-state reachability for WLCS. It should be possible to further apply the analytical model to research on the communication performance of e2e wireless channel.3. Performance analysis for e2e channel system with lossy communication of multi-hop wireless networks (E2E-CSLC). Based on probabilistic analysis for state reachability of WLCS, we further research on the performance of E2E-CSLC. Compared to WLCS, E2E-CSLC considers the impact of state of relay nodes among E2E-CSLC on the performance of E2E-CSLC. Based on the analytical model, we discuss some communication performances (e.g., e2e packet loss probability, e2e delay, and e2e throughput, etc.). E2E-CSLC is useful to design and improve e2e communication protocols.4. Performance analysis for (?, S)-bottleneck cell in large-scale wireless networks. The performance analysis of (?,S)-bottleneck cell is enormously difficult, primarily because of complex network interactions. We present a performance analytical model for (?,S)-bottleneck cell based on two-hierarchy QBD, and perform some analysis on the performances of (?, S)-bottleneck cell. To capture the essential aspects of (?,S)-bottleneck cell, we use two-hierarchy QBD. The general characters of (?,S)-bottleneck cell are govern by the first hierarchy QBD, while the characters of each flow are captured by the second hierarchy QBD. Based on the model, we present a methodology to derive some theoretic ranges for the transmission probability of a flow, and throughput. The dissertation focuses on modeling technique of the e2e communication channel systems in lossy networks where the underlying communication medium is faulty using QBD. Based on continuous-time and infinite-state QBD, it proposes a systemic methodology to analyze these systems, and derives some theoretic ranges for some performances. These modeling techniques overcome the limitations of existing analytical techniques and are helpful for designing and managing e2e communication in networks. These research results are the theorectic basis for improving the performances of unreliable networks and adaptive e2e communication protocols.