May 1st, 2001
Analyzing fast-growing networks involves considerable complexity, with the number of networked applications increasing exponentially. Existing techniques to analyze networks make use of network data collections, measurement and visualization to understand the state of networks. However, performance problems in networks may be caused by specific interactions from applications, which may or may not be running within a local network controlled by one network manager. This calls for two primary capabilities not addressed by traditional approaches. Network state information has to be first viewed and understood in the context of performance of applications. In addition, the information that can thus be obtained has to be shared with other remotely situated network managers, hence taking advantage of group collaboration to accelerate problem solving. Working towards this goal, this thesis proposes an extendable, unified and collaborative approach for analyzing networks.
The approach first proposes ubiquitous access to run active tests between machines on a network and basic network management queries, using the Simple Network Management Protocol (SNMP). A unified framework designed using a centralized server, allows simultaneous analysis through message passing between processed that monitor the results of a network test, and processed that monitor changes in a Management Information Base (MIB) entry of any network device. The framework also allows any new application written, to connect to the same centralized event servicer used, and send event messages concerning specific, possibly undesirable, conditions reached while execution.
The unified framework is complemented and merged with a collaborative framework to enable network managers to log-on to a group session and share the interfaces that control the network operations. Users can view results of operations initiated by others and also request control of interfaces that they see, to conduct their won tests on test-beds of other users in the session. The thesis demonstrates the result of such a unified and collaborative approach proposed by introducing a new type of network analysis tool called UCAN (Unified Collabratory for Analyzing Networks) implementing the capabilities described.
UCAN or (uCAN) employs a three-dimensional visualization tool to analyze network performance of applications. A generic, Configurable two-dimensional graphic visualizer is provided to monitor active tests and changes in MIB information using SNMP. uCAN also contributes utilities implemented as a reusable set of C++ classes built over the CAVERNsoft high performance networking toolkit and the FLTK user interface library. The utilities are designed to facilitate building new interfaces that can be shared in a group session as descried above. The goal is to help fit the new tools into the collabratory and give network administrators and users, a more complete view of an entire network.
Krishnaprasad, N. K., A Unified and Collaborative Approach for Analyzing Networks, Submitted as partial fulfillment of the requirements of the degree of Master of Science in Computer Science, Graduate College of the University of Illinois at Chicago, Chicago, IL, May 1st, 2001. http://www.evl.uic.edu/cavern/ucan/papers.html