The CORBA-based unified event management framework in multi-layer networks

Abstract

An efficient integrated event management is a great concern to a telecommunications industry. An integrated event management across multi-layer networks is much more complex than that across a specific layer network. Although there are several event management frameworks [Ericsson Research Canada March (2000); Integrated Network Management — Proceedings of the Fourth International Symposium on Integrated Network Management (1995) 290–301; Integrated Network Management — Proceedings of the Fourth International Symposium on Integrated Network Management (1995) 278–289], each of them is designed to apply only a specific single layer network. The frameworks are the cause of a lack of scalability and of the difficulties of integrated event management of multi-layer networks. On the other hand, CORBA Notification Service (NS) [OMG, CORBA Services: Common Object Service Specification March (1995); OMG, The Common Object Request Broker: Architecture and Specification July (1995)] is a promising event propagation platform to build an integrated event management for multi-layer networks, but its performance is not sufficient for bulk events in such networks.

This paper proposes the unified event management framework that can uniformly be applicable to multi-layer networks of ATM, Frame Relay (FR) and IP without any customization or modification. In order to define such a framework, this paper proposes a common network model and an event information model that is commonly applicable to the multi-layer networks, which identifies the meaning and relationship of information manipulated by the framework. This paper also suggests an event queue management scheme to improve the insufficient performance of CORBA NS stemmed mainly from its persistency property. Replacing this with the proposed queue management scheme and comparing the event management performance of the framework with that of CORBA NS, this paper will show that former gives a much higher event propagation performance than that of the latter.

Introduction

Network survivability is an issue of great concern to a telecommunications industry eager to deploy multi-layer networks of ATM, FR and IP. Because the loss of services at any layer network due to catastrophic or soft failures could be devastating and could result in significant revenue loss, many methods for maintaining network survivability are already proposed. To ensure the network survivability in telecommunication environment, there are two prominent management domains of event and fault. One is to rapidly detect network abnormalities including performance degradation and hard or soft failure, and another is to solve the abnormalities identified by the event management. Therefore, the network survivability depends fully on the scaleable and speedy event management scheme and the rapid restoration mechanism. However, the integrated event management across multi-layer networks is much more complex than that across a specific single layer network. Although there were several event management frameworks [1], [2], [3], each is designed to apply a specific layer network. They have been the cause of a lack of scalability and the difficulties of integration of multi-layer networks from the perspective of integrated event management. On the other hand, the increasing size and complexity of the multi-layer networks require high performance network management systems (NMSs) to be distributed and flexible. CORBA is considered to be a promising platform to build the distributed NMSs. The Object Management Group (OMG) Notification Service (NS) is an efficient and intelligent means of distributing information to many recipients in large and multi-layer network environments. The OMG has recently defined a standard NS that enhances the existing OMG Event Service by adding the following major capabilities, i.e. filter, priority and persistency [8], [9]. However, the performance of CORBA NS shows insufficient performance for propagating and managing the bulk events in multi-layer network environments, which is mainly caused by the overheads for filter, priority and persistency controls.

This paper proposes a unified event management framework that is uniformly applicable to ATM Virtual Path (VP), Virtual Channel (VC), IP and Frame Relay (FR) layer networks without any modification or customization. To define the unified event management framework, it is most necessary to know the generic network model that is not specific to a single layer network or a certain vendor's technology. So, this paper suggests a common network model that is uniformly applicable to multi-layer networks of ATM, FR and IP, and a generic event information model that is applicable to the proposed generic network model and that is described in the Rumbaugh's OMT model [13]. The event computational model interacting with CORBA NS is defined in terms of functional decomposition of distributed event management application. In that model, the event queuing scheme that is described with the CORBA Interface Definition Language (IDL) is proposed. Replacing the persistency property of CORBA NS with the scheme prominently improved the event propagation performance. The performance of the proposed framework was also evaluated under the high-speed information network (HSIN) of Korea Telecom for reality. The performance between the proposed framework using the queue-based and the CORBA NS persistency schemes were also analyzed.

Section 2 presents the common network model that can be applicable to multi-layer networks. Section 3 illustrates the event management framework in terms of information and computational modeling. Section 4 describes the implementation model using the CORBA NS and the results of the performance analysis. Section 5 discusses the merits and issues of the framework followed by concluding remarks.