I Introduction

 

Today, distributed systems based on the client/server paradigm have made their way into commercial IT infrastructures. The price of the flexibility gained is a more complex technical management of the computing environment. Efficient operation and administration requires an integrated management, in other words administration should be based on a single conceptual framework, namely a management architecture. This task has become even more complex through the recent introduction of additional management architectures (see e.g. [1], [2]); on the one hand, there is the well-known OSI management architecture that is primarily used in the telecommunications area; on the other hand, many IETF working groups are extending the scope of the Internet (SNMP) management architecture, i.e. applying it to the management of distributed environments. Additionally, CORBA [3] is becoming increasingly important for management applications. Whereas this architecture has not been developed specifically for management applications but to generally support communication and cooperation within distributed applications, it seems promising to use it for management purposes, too. Unfortunately, the current state of CORBA-based management of end systems and applications leads to an isolated management island. This produces a situation where not only the managed resources introduce a large amount of heterogeneity, but also the different standardized management frameworks. Therefore, bridging between the mentioned management architectures is an important research topic today. Our overall goal is to make integrated management feasible in an environment consisting of different management architectures.

The aim of the work described in this paper consists in helping to establish a ``native'' CORBA management environment that can also cope with managed systems in ``classical'' management environments that rely e.g., on the Internet management architecture: The management system should be ready for CORBA-based management but also take advantage of the functionality delivered by already existing management platforms. We achieve this by establishing a conceptually integrated management information base where management related information can be collected and evaluated independent of the originating management framework. It contains data of resources as well as events which have been raised somewhere ``out in the network''. The aim of our work lies in developing mechanisms which can be used for integrated but also distributed management of services, systems and networks.

In this paper, we will focus on two major integration problems, namely the transfer of events between both (CORBA and SNMP) worlds and the use of the management platform services by CORBA-based management applications. Therefore, we will explain the functionality of the Event Filtering and Logging services of the platform and a method to build a bridge for events flowing between CORBA and the platform. Events emitted by CORBA agents can be transparently routed via the bridge into the event services of the management platform, making use of its filtering, displaying and storage capabilities. Platform filters can be set up using CORBA interfaces. By using regular features of the platform, these events may trigger management actions. We will also explain the functionality of the platform Topology Services, which manage topological relationships between managed objects and show how these generic services can be used in a straightforward manner to visualize the relationships between CORBA-based services and objects. For this visualization, we have developed a simple relationship model tailored for the LEO/MEO satellite environment described in section II. To achieve the integration of events and access to platform services, we made use of the features of the CORBAservices [4] by building ``wrapper'' objects which open the programming interfaces of the platform and its services to objects anywhere in the CORBA environment.

The structure of the paper is as follows: Section II introduces a real-life management scenario, namely the CORBA-based management of LEO/MEO satellite constellations. They are an extreme example of a distributed environment and have management requirements similar to other distributed systems. We believe that distributed environments should also be managed in a distributed manner and therefore propose CORBA as management architecture. As currently no CORBA-compliant management platforms exist on the market, there is a strong need for achieving interoperability between well-established SNMP-based management platforms and the managed nodes represented by CORBA agents. The third section of the paper presents the different possibilities for bridging the gaps between heterogeneous management frameworks and gives reasons why we decided to perform the integration on the side of the managing system, i.e. making an SNMP-based management platform suitable for CORBA. It also discusses the different platform-side integration alternatives and gives an overview over the services provided by the platform that are most useful for our purpose. The techniques how this integration can be achieved will be described in section IV which presents the results of a study and a trial implementation of a CORBA-compliant management platform which integrates a "conventional", i.e. SNMP-based network management platform product (IBM NetView for AIX) with an Object Request Broker. Section V concludes the paper and gives an overview of further steps.