Mobile Agent Approach to Congestion Control in Heterogeneous Networks
One of the motivations to study the behavior of the Internet is to find out the best way to maintain the relative stability of the global network. This leads into the investigations of events that impair the performance of the system such as congestion that occurs whenever the demand for resources exceed the available capacity. When the congestion is left uncontrolled the performance of the whole system degrades through severe delays, lost packets, and even a complete shutdown of the network. Hence, congestion management through monitoring, detection and control is necessary in order to sustain acceptable levels of network performance and this may be done via the transport protocols. Consequently, many modifications of the original TCP protocol have been implemented to manage the control. On the other hand, unlike TCP, UDP has no knowledge of congestion whatsoever and hence unresponsive to the network problems.The work explores the possibility to influence and modify the unresponsive behavior of UDP or similar protocols via the mobile agent paradigm. The autonomous entities are able to migrate across the network and sense the state of the network and when needed tame the intensity of UDP or alike flows to prevent congestion. The proposed model is termed the Combined Model for Congestion Control (CM4CC) and has two different objectives. The first one is to employ the host centric or end-to-end (E2E) congestion control mechanisms for the TCP flows; the second one is to invoke the mobile agent paradigm to manage the non-TCP (or UDP) traffic. Both mechanisms must work together to avoid congestion. When it eventually appears, they have to assist the network in speedy recovery and return to the normal mode of operation. The validity of the CM4CC has been verified through numerous simulation scenarios using the Optimized Network Engineering Tool (OPNET). The results provide the basis for an environment that makes possible the coexistence of responsive and unresponsive flows.
Source Type:Doctoral Dissertation
Keywords:SOCIAL SCIENCES; Statistics, computer and systems science; Informatics, computer and systems science; Computer and systems science; data- och systemvetenskap; Computer and Systems Sciences
Date of Publication:01/01/2008