OBJECT BASED QOS MANAGEMENT FRAMEWORK

Pradeep Ray,

Department of Computing

University of Western Sydney, Nepean

Kingswood, NSW 2747

Australia

Tel:+61 2 9685 9251, Fax: +61 2 9685 9245, email: pradeep@st.nepean.uws.edu.au

ABSTRACT

Multimedia applications, such as videoconferencing have varying performance requirements and these are best expressed as Quality of Service (QoS) parameters. There is a need for a multilevel hierarchical QoS management framework to address the requirements of emerging multimedia services in global telecommunication environment (e.g., supporting mobile and static users) in view of the growing complexity of modern information networks. Recent developments in network and systems management suggest the use of integrated management platforms for a wide variety of management tasks. This paper presents a multilevel framework for QoS management in a global corporate network, consisting of heterogeneous elements, such as static and mobile communication segments. This is based on TINA connection management framework. Partial implementation of this framework has been undertaken, using a commercially available CORBA platform.

1.INTRODUCTION

While perceptions of QoS vary from user to user and from application to application, these requirements need to be translated (mapped) to a finite set of network and system resource requirements. QoS Management is thus concerned with the specification, mapping, monitoring and control of QoS in a distributed application [Hafid 95]. In a large corporate organisation the end user is served by the internal system support people who are, in turn supported by value added service providers and talcum carriers. There are cost and service quality issues at each level of interface.

Hence there is a need for abstract modelling of network and system resources for QoS mgmt. This paper presents a QoS management architecture based on ISO Open Distributed Processing (ODP), Bellcore initiated Telecommunication Information Networking Architecture (TINA) and the Object Management Group (OMG) Common Object Request Broker Architecture (CORBA).

There is now a trend to develop management applications on integrated management platforms (e.g., SunNetManager, Cabletron SPECTRUM, HP OpenView etc.) that support multivendor management applications and heterogeneous ``instrumentation'' [Ray 93]. Therefore, this paper describes a hierarchical QoS management on top of the integrated management framework discussed in this section. A partial implementation was undertaken, using Digital ObjectBroker CORBA platform (not discussed here due to space limitations).

2. QOS MANAGEMENT

QoS management is concerned with the support of a QoS-bound application (QoS parameters to be maintained within certain limits) between two or more users. A QoS management process requires remote cooperation of system users (e.g. medical professionals in a Medical Imaging application), system support personnel, value added service providers, telecom carriers, network equipment vendors.

Therefore, it may be necessary to present different abstract views of service configuration. Hence there is a need for modelling QoS management at several levels of abstraction starting from the highest level.

2.1 QOS MANAGEMENT PROBLEMS

The problems of QoS Management can be broadly classified as:

1. QoS Specification concerned with varying QoS requirements for applications and no standard method (set of parameters) for the specification of QoS for all applications, and unpredictable variations in user perceptions and hence the difficulty in assigning any values to these parameters

2. Mapping to Distributed System Resources concerned with the user specified parameters (e.g., good resolution, colour, quick response etc.), need to mapped to system technical parameters, such as throughput, delay etc. that require changes beyond the control of an end-user, such as: communication channel bandwidth, communication protocols, and operating system resource management

3. Monitoring and Control issues such as end-to-end QoS management requiring centralised monitoring and control of QoS [Seneviratne 94]- This is difficult because a distributed application covers a number of independently administered network domain and systems. Each domain has heterogeneous equipment requiring different management knowledge, and widely varying unpredictable network and system performance (e.g. mobile wireless environment). Most of the services are based on the Internet where the only QoS one can guarantee, is ``Best-Effort''.

2.2. WORKFLOW MODELLING OF QOS MANAGEMENT


Fig.1 shows the QoS Management application using a workflow model where each loop shows an independent task with customer on the left and performer on the right of the loop [Action 95]. This notation provides us with a good representation of organisational processes. The QoS Management task is divided into three sub-tasks: Resource Reservation, Protocol Selection and QoS Monitoring. A requirement analysis of QoS Management shows that the system is involved in a ternary association with the following participants:.

1.User (specifies the service, expects response from manager according to rules)

2. Network Manager (communicates with the user and telecom provider, translates user needs to appropriate specs for resources required, protocol stack and monitoring MIB,)

3.Telecom Provider (reserves resources, maintains service-reports in case of failure)

2.3. ISO ODP COMPUTATIONAL VIEWPOINT OF QOS MANAGEMENT

[Znaty 94] describes QoS Management for an FDDI system from ODP viewpoints. This paper analyses the problem from the perspective of end-to-end QoS management, using the ODP computational viewpoint [ISO ODP].

In this case, an explicit multiparty binding is established through object QOSCOMPUTATION&CONTROL (Fig 2). The interfaces are: User (operational,stream), QoS Monitor (operational, signal), QoS Resource Manager (operational), and Protocol Server (operational).

In addition to these, the QOS-COMPUTATION&CONTROL OBJECT invokes management operations through OSIMANAGER object. This uses OSIAGENT object to access systems management information. Both OSI and SNMP management use this manager-model for realizing the ``instrumentation'' (accessing management information through MIBs). The multiparty binding object maintains QoS management information for use by itself and other objects connected to it.

While the encapsulated functions of QOSMONITOR and QOS-RSRCMNGR are straightforward, USER(S) incorporate intelligence required for the specification and mapping of user QoS parameters.

Although ODP provides a set of mechanisms for the description of a distributed system from different perspectives, it is felt that there is a lack of an explicit mechanism to describe QoS at different levels in a network/service (e.g., WAN, LAN, Device).

Formal specification of these objects can be achieved using CORBA Interface Definition language (IDL). However, OSI managed object specification (at instrumentation level) is based on ISO/ITU-TS International Standard Guidelines for the Definition for Managed Objects (GDMO) [ISO 10165]. Internet SNMP is not based on object-oriented concepts. Therefore, management platform vendors use various proprietary object models (other than CORBA) to encapsulate management information. In this project, we used a commercially available CORBA platform (Digital ObjectBroker), in conjunction with Cabletron SPECTRUM network management station for partial implementation (not discussed in this paper due to space constraints).

3. LAYERED QOS MANAGEMENT

The QoS management problem will now be examined in light of the management architecture proposed in TINA [Fuente 94], developed by an international consortium of telecommunication organisations, called TINA-C. TINA defines Connection Management as a new management function in addition to OSI functions Fault, Configuration, Accounting, Performance and Security (FCAPS). This is useful in the Quality of Service (QoS) management in distributed multimedia applications.

3.1. TINA CONNECTION MANAGEMENT

The OSI Reference Model [ISO-7498] combines all networking functions, such as routing, connection establishment, congestion control, internetworking and so on into a single layer (network layer). This is inadequate in view of the recent advances in distributed computing and the need for their utilisation in the above networking functions. TINA proposes to handle this as part of the Connection Management function. Network resources in TINA are seen grouped in a set of layer networks as in Telecommunication Managed Networks (TMN) standards [M.3010]. Characteristics of different TINA layered networks are:

Each layer network is managed individually

Each layer network operations are performed independently of other layer networks

A layer network can access service of another layer through client-server relationships

A layer network may be split into several domain

Management requirements are first specified at the service level and then translated to abstract resource level. Finally these are mapped onto the network element level. In order to manage network resources independent of specific devices, a number of resource management fragments are being defined (e.g., fault management, resource configuration management, connectivity etc.).

TINA Connection Management is achieved by providing multiple layers of objects given below [Fuente 94]:

Communications Session Manager(CSM)

is the top level object that provides for setting up, maintenance and release of logical connections. The specification refers to computational object interfaces instead of addressable points in the telecommunication network.

Connection Coordinator(CC)

provides interconnection of addressable termination points of telecommunication networks. The specification consists of termination addresses and connection characteristics, such as QoS parameters (independent of underlying network technology)

Connection Performer (CP) and Layer Network Domain Coordinator (LNDC)

provide interconnection of termination points of the subnetwork. There are a number of types of different connection performers depending on the technology used (e.g., ATM, cellular mobile etc.). There is a coordinator for each domain in a network layer (LNDC) which receives requests for connectivity in a layer network, and has federation capabilities with other LNDC of other domains in the layer network.

3.2. QOS MANAGEMENT IN A GLOBAL ENVIRONMENT

QoS Management in a large network is a complex problem. Mobility adds a new dimension to this problem [Imielinski 93]. The TINA Connection Management Architecture appears quite useful from this point of view. The modelling was done for a distributed environment where medical professionals need to be mobile and cooperate remotely over a heterogeneous network environment. This helped us model different levels of Resource Managers required in a network with static and mobile wireless segments [Ray 95b].

Management of mobile communication services require the segregation of control between the mobile and static segments of the network since the mechanisms for control vary substantially. For example, QoS in mobile networks vary widely depending on atmospheric conditions for wireless links unlike in static optical fibre links. Media access protocols are different for mobile and static segments of a network. On the other hand, it is necessary to have a higher level session manager that would support QoS on an end-to-end basis. TINA proposes a connection management scheme that would help in the integrated management of QoS in a complex network with mobile and static parts.

As discussed earlier, QoS management is a difficult problem since it affects all layers of the network. Fig.3 shows QoS Management as a multi-level network-wide system. This helps model the components of a QoS management system at various levels (e.g., corporate network, LAN, Network element etc.).

In this model the QoS problem is illustrated in a Mobile Multimedia Environment, in terms of two independent client-server interactions:

Client-Server in Static segment

Client-Server in the Mobile segment

While the diagram at the top shows a high-level object model of the application, the lower diagram shows different building-blocks required to implement the scheme. One can now model the QoS Management computational objects (Fig 2) as instances of different blocks of TINA Connection Management as follows:

Protocol Server as a Resource Configuration Manager

QoS Manager as the Connection Coordinator in each layer

Resource Managers at different levels as LNDC and CPs

Fig 3 shows a picture encompassing both the static and mobile parts of the system. Integrated QoS Management can now be achieved by

Splitting QoS Management for Mobile and Static part of the network at the Mobile Base Station; this involves

a mobile client-server [Ray 95a]

existing management blocks work upto the edge of the static/mobile networks

A new scheme for QoS MNGR and RSRC MNGR for the mobile environment; this involves the framework proposed in [Ray 95b]

4. CONCLUSION AND FUTURE WORK

This paper has presented a discussion on Quality of Service (QoS) Management at computational level on the basis of ISO ODP and TINA frameworks. This started with a description of a QoS management application using workflow modelling. Then there was the ODP computational viewpoint description of a QoS management framework on the basis of integrated management framework described in the beginning. In view of the complex multilevel structure of QoS management application, it was felt necessary to use richer modelling semantics. This led to the study of TINA and the modelling of QoS management framework on the basis of TINA Connection Management recommendations.

ACKNOWLEDGEMENT

The work reported in this paper has been funded in part by the Cooperative Research Centres Program through the Department of the Prime Minister and Cabinet of the Commonwealth Government of Australia.

REFERENCES

[Action 95], Action Technologies Inc., USA, ``Action Workflow'', http://www.actiontech.com/action/

[Fuente 94], Luis A. Fuente et al., ``The TINA-C Management Architecture'', 5th IFIP/IEEE International Workshop on Distributed Systems Operations and Management (DSOM'94), Toulouse, France, October 1994.

[Guha 95], Deb Guha et al., ``TINA Overview'', International Conference TINA'95, Melbourne, February 1995

[Hafid 95], Abdelhakim Hafid et al., ``An Approach to Quality of Service Management for Distributed Multimedia Applications'', Proceedings of International Conference on Open Distributed Processing (ICODP'95), February, 1995

[Imielinski 93], T. Imielinski et al., ``Data Management for Mobile Computing'', SIGMOD RECORD, Vol. 22, No.1, March 1993

[ISO 7498-4], ISO-7498-4/ITU-TS-X.200, ``OSI Reference Model: Management Framework'', ISO/ITU-TS International Standards 1988

[ISO 9596], ISO-9596/ITU-TS-X.711, ``OSI Reference Model: CMIP Specification'', ISO/ITU-TS DIS' 1989

[ISO QoS}, ISO CD, ``QoS Basic Framework'', January 1995

[M.3010], CCITT M.3010, ``Principles For A Telecommunications Management Network, ITU-T Draft Recommendation

[Ray 93],Pradeep Ray et al., ``Integrated Management: Emerging Trends and Future Challenges'', 8th International Joint Workshop on Computer Communication (JWCC-8), Taipei, Taiwan, December 1993

[Ray 95a], Pradeep Ray et al., ``Integrated Management in a Mobile Environment: A TINA Perspective'', International Conference TINA'95, Melbourne, Australia, Feb. 1995.

[Ray 95b], Pradeep Ray et al., ``QoS Management in a Mobile Multimedia Environment'', 7th IEEE Workshop on LAN/MAN, Marathon, Florida, USA, March 1995.

[Seneviratne 94], A. Seneviratne, M. Fry et al., ``Quality of Service Management for Distributed Multimedia Applications'', IEEE Phoenix Conference on Computer and Communication, Phoenix, Arizona, USA, 1994

[Znaty 94], Simon Znaty et al., ``ODP Viewpoints of QoS Management Application'', 5th IFIP/IEEE International Workshop on Distributed Systems Operations and Management (DSOM'94), Toulouse France, October 1994

AUTHOR TRACK RECORD

Pradeep Ray is a member of the academic staff in Computing at the University of Western Sydney, Nepean, Australia. He has a Masters in Electrical Engg. from the Indian Institute of Technology, Kanpur, India, and he has recently completed PhD in Computing Sciences from the University of Technology, Sydney. He has more than ten years' experience in the international telecommunication industry, and academic/research establishments. His research interests include network and systems management, distributed/cooperative systems, mobile computing and multimedia enterprise networks. Pradeep is organising the 8th IFIP/IEEE International Workshop on Distributed Systems Operations and Management (DSOM'97) as the Program Chair.


Last Modified: 02:26pm , April 22, 1997