3.3 Relationships of MSs to the System and Network Infrastructure

  MSs usually run on general-purpose operating systems like UNIX or WindowsNT and are therefore sensitive to errors that occur in the underlying operation system and the communication infrastructure provided by networked services (e.g., the Domain Name System (DNS), the Network File System (NFS) and the Network Information System (NIS)) and protocols. Consequently, commercial MS implementations like the one described in [5] dedicate a high degree of their management instrumentation to the monitoring of the underlying system. While this may be necessary, in our opinion it is not the primary concern of a MIB for MSs to control not only the MS itself but also the underlying environment. Thus, we restrict ourselves to the instrumentation of the MS itself and consider the instrumentation of operating systems (CPU usage count, user quota, disk and paging space etc.), networks (connectivity checks, timeout errors, network buffer size, packet and frame errors) and underlying networked services like DNS, NFS and NIS as out of scope for this paper. Nevertheless, we have designed object models and implemented the corresponding agents for UNIX systems and networked services. Readers interested in these agents are referred to [16]. The development of these agents has been done according to the approach described in section 4, too. On the other hand, operating system configuration tools like IBM AIX SMIT provide a vast amount of static management information pertaining to MSs such as the names and versions of the software modules, their installation paths, their prerequisites and dependencies. As these tools are accessible through APIs we were able to retrieve this management information.

We consider management itself as a distributed application; the consequence that the distributed application ``management'' itself needs to be managed has already been motivated in section 1. Another implication is that we can apply already defined and standardized models for (general-purpose) distributed applications like the Reference Model of Open Distributed Processing (RM-ODP) to the management of MSs. We will show in the following section how we used concepts from RM-ODP to obtain generic management object classes for application management; we will use these as base classes for building the inheritance hierarchy of our object model.