The advent of object-oriented distributed processing models has brought forward a third candidate, CORBA, originally not targeted specifically at management tasks, but in many respects nevertheless suitable for managing both local and wide area networks.
CORBA is more powerful than SNMP and less complex than CMIP. Its affiliation with C++, a widely used programming language, lends it to immediate use by a huge number of programmers and allows them to introduce distribution into their programs without a too drastic change of philosophy.
It is therefore assumed that CORBA will become important in the network and systems management domain as well as in the distributed systems domain. To be more precise, CORBA will be used to implement management applications (managers) and managed entities (agents).
The author assumes that, in the near future, all models will have to coexist, because investments in older models have been made, because CORBA may not yet be ready for certain specialized management tasks (e.g. embedded management agents), or simply for political reasons.
Therefore, the need arises to manage a system written in one model from a system written in a different model, for example to manage an OSI agent using a CORBA-based manager. Assuming that CORBA will achieve broad acceptance in the management world, it would be desirable to be able to manage other models transparently from CORBA. The benefits would be investment protection of existing managed entities, the opening of the SNMP- and CMIP-dominated management world to management-inexperienced (CORBA) programmers, and the unification of management in one common, simple, and yet powerful model.
The goal of this work is to examine how CORBA can be used for network and systems management. The focus is on the client side; that is, how CORBA can be used to implement management applications that access managed entities, rather than how CORBA can be used to implement managed entities.
Work in this area already exists; an overview will be given and it will be shown that most work focuses on compile-time static translation of one model to another. This generates a number of problems.
Therefore a dynamic runtime-based approach is proposed that eliminates these problems and has several advantages over static approaches. The proposed model is a combination of a generic object model, metadata, and adapters that convert between the generic and specific target models. Although the elements of the proposed models are established and well known, their combination and their application to the domain of management is novel and makes for the issues of interest in this thesis.