 |
 |
 |
 |
|
|
|
|
Operations, Administration, Maintenance functions are increasingly moving down toward network elements (NEs), making them self-sufficient. Networks need to be capable of self-regulation with respect to traffic loads and self-healing or survivable with respect of failures. Functionally, networks will become increasingly distributed, by network elements taking control of their own management.
The management of today's and future networked environments requires far more than network management in the traditional sense, where the managed entities are physical lines, nodes, switching equipment, network connections, and so on. Management of distributed applications and user oriented services adds a new dimension to the problem. The management functions pertaining to distributed systems and applications aspects often are disconnected from the underlying network management. In order to archive a truly integrated management of network and distributed system aspects, both of these disparities need to be addressed. The network management system have to be able to manage a broad range of networks of different technologies. In ICM we decided to use ATM (Asynchronous Transfer Mode) network ETB (EXPLOIT Testbed). The ATM technology is leaving field trials and ATM networks start appearing, first as LANs and MANs, forming the basis of Customer Premises Networks that will connect to public B-ISDN. Closer interaction between the real-time network control mechanisms and network management is being sought to provide more efficiently guarantied QoS. This QoS cannot be satisfied alone by the basic networking technology (switches, adapters, media access control protocols). Integrated management of the system and network resources is required to achieve control over quality of service end-to-end.
It is clear that standards are highly important to being able to operate and manage networks built of heterogeneous components. Standards do not reflect correctly the present situation and trends in open network management. SNMP can be seen especially in LAN-based environments and the Internet community in general. This trend is understandable, since the OSI standards become stable only after SNMP had already been deployed quite widely in the Internet, and are significantly more complex. The OSI (CMIP) standards provide a powerful object-oriented management information model, support flexible distribution of management functionality and provide common system management functions which together make them well suited for the management of large heterogeneous networked environments and distributed applications. A very important step in deploying the OSI standards is to specify a common base of managed object classes. GDMO is the standard notation developed for this purpose. The main work on specifying object classes for network management is carried under the auspices of the Network Management Forum, in the OMNIpoint program.
It is clear that standards are highly important to being able to operate and manage networks built of heterogeneous components. Standards do not reflect correctly the present situation and trends in open network management. SNMP can be seen especially in LAN-based environments and the Internet community in general. This trend is understandable, since the OSI standards become stable only after SNMP had already been deployed quite widely in the Internet, and are significantly more complex. The OSI (CMIP) standards provide a powerful object-oriented management information model, support flexible distribution of management functionality and provide common system management functions which together make them will suited for the management of large heterogeneous networked environments and distributed applications. A very important step in deploying the OSI standards is to specify a common base of managed object classes. GDMO is the standard notation developed for this purpose. The main work on specifying object classes for network management is carried under the auspices of the Network Management Forum, in the OMNIpoint program.
Network management and traffic control system are currently used in decision support mode to manage networks. As traffic control expert systems become closely integrated with alarm surveillance systems and are tested sufficiently to verify the accuracy of their decisions. At the same time, as the systems are extended with more knowledge about operations and alternative controls, they will more closely support human experts in making complex decisions. Then there is also question as to how the Artificial Intelligence systems may be deployed.
Network management system is characterised by two main properties: the structure of management information and the management operations supported by the underlying protocol. With the purpose to manage real ATM network (ETB testbed) ICM implemented CMIP agents on each Subsystem Control (SC) of the ETB. Agent is responsible for the interface to the software resources under the control of the TMN. This introduces NE(etb) into the physical architecture. Each, NE(etb)will contain a NEF, QAF(qx), MF and provide a Q3 interface to the NELM. The Q3 interface is the TMN standard interface that provides to an Operation System (OS) the standardised view of the network elements to which it directly interfaces. The Q3 interface consists essentially of two components: a CMIP protocol stack and the information model of the managed network (ETB) element. The information model is defined in terms of managed objects (MOs). Management objects are data images of the physical and logical resources that implement the managed network elements. The definition of these MOs specific to a particular network element is not much supported by current standards. The managed objects are formally specified using the Guidelines for the Definition of Managed Objects (GDMO) which is a formal ASN.1 template specification language. Some concepts from the object-oriented approach enrich the OSI model like: -Inheritance, which shows the inheritance relationship among the managed object classes. This relationship provides the reusability function; - Containment, which represents the aggregation relationship among instances of classes. It models a "kind of" relationship to denote which managed objects are "part of" another managed object. The architecture of the MIB identifies the following elements:
The QAF consists of two interfaces: A Q3 interface to the TMN system, realised through CMIS/P, and a proprietary M interface to the network. The current implementation assumes that there is one QAF for the whole network, which is not embedded at any node. The Q3 interface is implemented on the OSI stack, where the upper four layers can be generic, while the lower three layers will be depending on the physical medium over which the Q3 interface is implemented. The Q3 interface is implemented using OSIMIS on the ISODE OSI stack. The information kept at the QAF is an object-oriented representation of the information that the TMN requires to from the network. The TMN observes the required data by polling. No processing is carried out in the QAF, apart from the translation of the CMIP messages to/from M-interface messages. The M-interface messages are forwarded from the network to the QAF and vice versa through a socket-based communication mechanism.
ICM has implemented Q-adaptors to commercial ATM equipment in order to widen the scope of its work over the ATM Pilot.
One of the key aspects of the ICM project is that the research concepts which are proposed are validated through an implementation and testing process. Design of the MIB has been extensively investigated and the final proposal is the result of an in depth iterative process between implementation and modelling. QAF MIB is simple enough to be easily implemented and also contain enough information to be used by the management applications. The object oriented approach used within GDMO enhances reusability
It is intended to submit the ATM MIB (GDMO) directly to the standards, once consensus has been achieved via STGs within RACE. ICM partners (who are manufactures) will be able to incorporate the work done on the ATM MIB within products in the future and would have an advantage (in time) over competitors. The ICM project will reuse this work within Phase 4 and it is expected that the MIBs of other ATM networks will be used by ICM. The result of our work will be submitted to R2083 as CFSs contributions, and ICM will also make the results available to other projects working in this area.