2 Management model of distributed applications

 The question ``what are the characteristics of distributed applications from a management point of view?'' is a very current research topic. It is addressed in various recent publications (see e.g. (Pell et al., 1995; Schade et al., 1996; W.Hong et al., 1995)). Compared to other management areas, there is still a lack of common understanding of this issue (see e.g. the ongoing discussion related to the development of Application MIBs (applMib and sysApplMib) for the Internet management architecture). In order to describe these characteristics, we first need a framework for management objects. This framework must allow us to describe relevant aspects of any distributed application in a way that is suitable as a basis for all areas of management. For example, it has to allow the specification of relevant aspects that arise e.g. from software management and distribution on the one hand and from the need to monitor the status of processes and their ability to communicate etc. on the other hand.

Today, there seems to exist only one standard model that is comprehensive enough to address all management issues -- the RM-ODP. It is not only standardized but also of growing practical importance due to, among others, the cooperation between the OMG and the ISO. Therefore, we propose to take it as a basis for the definition of the necessary framework for management objects. Our approach consists essentially of three steps (see also Figure 2):

  

Figure 2: Derivation of GAMOCs

1.

Defining MOCs for applications: First of all, we have to define the Managed Object Classes themselves, i.e. we have to identify components of distributed applications that are relevant for management purposes. This is done by an analysis of the concepts provided by the ODP viewpoint languages.

2.

Defining and assigning management-relevant information to the MOCs defined in 1.: This second step is a top-down approach that analyzes requirements of management functional areas, i.e. it addresses the question ``what information on resources do we need for management purposes?''. The identified information is then assigned to suitable MOCs of the first step.

3.

Ensuring feasibility: In the third step, the top-down approach is complemented by a bottom-up approach based on an analysis of information that is in general available from these resources, and thereby addresses the question ``what can reasonably be assumed in general about resources of that class?''

The main benefit of our approach is the wide applicability of the resulting management information to the majority of applications. It takes advantage of the fact that the ODP concepts have been defined with generality as a major design goal. It therefore should foster integrated management of distributed applications. In the following, these mentioned three steps are described in more detail and their application will also be sketched out.