Decision Support For The Provision of Services
Integrating Knowledge-Based and Optimisation Techniques for Resource Assignment.

 

 

Mark Tierney
E-mail: mt@broadcom.ie

Robert Davison
BT Labs, Martlesham Heath,
Suffolk, UK.
 

Abstract


The provision of telecommunications services has changed drastically over the last decade, due to major changes in technology, the market-place and customer expectations. The days when customers were only able to choose from a limited range of services offered by a single organisation are gone. Enormous increases in transmission capacity combined with increased network intelligence has led to an increase in the variety of services offered. For any particular customer request, there are a range of possibilities that a service provider can offer. The solution required will be the one which satisfies both the service provider and customer goals with respect to business objectives and quality of service.

Those service providers who develop efficient, responsive and rapid service provisioning processes will be well positioned to succeed. Traditional computer-based support systems are already being deployed to achieve this. This paper describes how the application and combination of decision support technology, such as artificial intelligence and operations research, allows DSS to be developed that provide more effective support to cope better with the changing telecommunications environment. Particular emphasis is placed on resource assignment activities.

This paper presents results and experiences gained in the DESSERT (R2021) project. A strength of the DESSERT project has been the successful combination of diverse advanced information processing techniques to provide decision support for the provision of services.

1. Introduction

The continuing expansion, performance and flexibility of the telecommunications infrastructure are providing a new and important focus to the forging of telecommunications strategy world-wide. Large and small business users are demanding universal access to a flexible and reliable telecommunications network, while domestic subscribers are led to believe that on-demand entertainment and information services are just around the corner. For service providers, the provision of high integrity, flexible services is increasingly seen as a major source of competitive advantage, at both a national and global level [1]. In order to meet this demand, fundamental changes will be necessary in the structure and supply of telecommunication networks, with traditional boundaries between switching and access, junction and trunk transmission topologies likely to disappear in the process.

Technology trends suggest a move from the complex, active, heterogeneous, copper-based networks of today towards simple, passive, homogeneous optical networks of the future [2]. Bandwidth and communication distance will cease to become the most important factors and functionality will become crucial. Provisioning will become more concerned with configuring the network access point and setting up access to appropriate pieces of software and less concerned with network resources or capacity. In addition, the customer requirements (both technical and non-technical) will have to be well understood in conjunction with key knowledge about the state and deployment of networks and services. Indeed, service providers will need an effective provisioning process if they are to survive [3].

Using computer based support systems service providers will be able to meet customer demands more effectively, more economically and more dynamically. To date, there have been few computer support tools to aid in the provisioning process. In fact, a substantial amount of provisioning is traditionally 'pen & paper' based and very time consuming. Sophisticated software needs to be developed that will assist personnel in their decision making process and enable computers to work in partnership with human experts. Powerful Decision Support Systems (DSS) are prime candidates to help personnel cope with increasingly complex engineering problems associated with telecommunications service provision. Such DSSs are software modules that will help automate the decisions faced in provisioning. The objective is to help the user make non-trivial decisions, understand deciding factors, understand trade-offs and provide computational support where possible.

This paper presents the results and experiences gained in the DESSERT1 (R2021) project. A strength of the DESSERT project has been the successful combination of diverse Advanced Information Processing (AIP) techniques (namely Knowledge-Based Systems and Operations Research) to provide decision support for the provisioning of services.

Three prototype Decision Support Systems (DSSs) have been developed within DESSERT which aim to address different stages of provisioning:

The project has also developed a DSS Toolkit and Methodology to support the rapid and efficient production of new provisioning DSSs using reusable software components [6].

In what follows we report on the approach and experiences gained in developing the Generation and Selection of Alternative Configurations (GSAC) DSS prototype. Section 2 describes the motivation for decision support. The approach and overall design are presented in section 3 followed by a discussion of the prototype in section 4. Finally, section 5 draws conclusions from our experiences with the GSAC DSS prototype.

2. Why Decision Support?

The provisioning of services requires a combination of human judgement and computing power. Human judgement is needed to select a solution that best meets the customers requirements while meeting the business objectives of the service provider. Computing power is needed to deal with the large quantities of information needed to manage today’s networks and services. DSSs can play the role of interpreter for this information, by taking raw data and turning it into meaningful information for the user to work on. DSSs should be supportive co-workers, which analyse and process information, presenting the user with suggested options, solutions and strategies. DSSs couple the intellectual resources of individuals with the capabilities of computer based support systems for management decision makers who deal with semi-structured problems [7].

In an increasingly complex domain, such as service provisioning, it is necessary to utilise the users ability to the full, and to supplement this with powerful computing tools which support the user in a decision making process. However, solutions to problems in such a domain will require an integration of different types of knowledge. For example, certain aspects of the problem may be solved through the application of well-known optimisation algorithms. Other parts of the problem may also lead to complexities2 that are beyond the scope of such techniques, requiring approaches that are more heuristic than algorithmic in nature. Moreover, effective DSSs must also possess a significant body of linguistic and presentation knowledge to interact with the user, and reasoning knowledge to exhibit intelligent behaviour.

Successful DSSs are likely to be a hybrid of a number of techniques together. The approach DESSERT has taken to achieve decision support for service provisioning, is to combine diverse Advanced Information Processing (AIP) techniques under the following classification:

The objective of such an approach, enables the inter-working of a wide variety of well understood, newer state of the art techniques within a single unifying framework. The fields of OR and KBS both address problem solving and decision making, but yet they have historically progressed in isolation [10]. KBS techniques have emphasised the use of strategies associated with human behaviour whilst OR has emphasised optimal strategies using mathematics. Both fields provide methods that can be applied in the service provisioning domain. It is envisaged that the synergy of these techniques will provide powerful decision support. The main benefits are facilitating and improving the quality of decision making, by reducing the information overload and by augmenting the cognitive limitations and rationality bounds of decision makers.

3. A Decision Support Approach to Service & Network Resource Assignment

New network and service technologies are being quickly deployed. Some of the issues that will need to be addressed by service providers are: In organising a service for a customer a number of service and network resources must be assigned to offer the appropriate Quality of Service (QoS). There are few computer support systems to aid in this process. In fact, a substantial amount of service provisioning has traditionally been a 'pen & paper' exercise, and has consumed much time and personnel resources. The computer solutions that do exist are generally constructed based on a single type of technique and are also dedicated to support the assignment of one kind of network resource (e.g. ISDN) only. The effort involved in assigning these resources will be significant, considering the scale and diversity of networks, and the multiplicity of services that are expected in the predicted growth of telecommunication services.

The availability of Decision Support Systems (DSSs) will support telecommunication personnel and increase their effectiveness and flexibility in being enabled to work with diverse resources of many different network technologies. The result being an integrated and flexible solution where many kinds of networks are catered for. This will reduce the unit cost of resource assignment, increase labour flexibility and will assist in reducing the delivery period for a service. DSSs can also more optimally configure networks and thus reduce expenditure on capital resources.

The expected benefits of decision support in provisioning are:

3.1 The GSAC Decision Support Approach

The GSAC DSS supports its user in performing resource assignment for the provisioning of services. Performing resource assignment requires a mixture of activities some of which are computationally intensive (e.g. optimising bandwidth usage) and some of which are knowledge intensive (e.g. ensuring that a set of resources meets a security requirement).

Resource assignment is the process of identifying service and network resources that will be needed for a particular set of customer services. For example, tasks such as designating connection points, routing cables, installing customer premises interfaces, selecting a port access to the local switching equipment and instantiating the customer with the requested class of service be part of this process. In order to do this, the request for the service needs to be examined in depth in order to determine

It is likely that there will be more than one possible solution, in which case the best solution will be chosen where 'best' is judged on a mixture of quantitative criteria (such as installation cost, operation cost, etc.) and qualitative criteria (such as security and extensibility). The solution required is the one which satisfies both the service provider and customer goals with respect to the business objectives and quality of service, respectively.

4. The GSAC System

In the design of any Decision Support System (DSS) it is important to consider which aspects will be performed by the USER and which by the DSS. Figure 4.1 gives an overview of the components and roles in the GSAC DSS.

It consists typically of a human decision maker (the user) and a computer based system, with both being viewed as resources in the decision making process and assuming different roles throughout the interaction process. For instance, the role played by the user of the DSS consists primarily in providing:

Figure 4.1: Overview of components and roles played during the decision making process in the GSAC Decision Support System (DSS).

In terms of GSAC this requires the user to perform three roles that are not fully amenable to automation:

On the other side, the role of the computer based system (the GSAC DSS) can be reduced to two main components, implementing support for the user in two areas:

4.1 GSAC Design

The design of the GSAC DSS follows the overall DSS application architecture illustrated in Figure 4.2. The fundamental building block within the GSAC DSS is the 'tool'. Each tool either automates a specific service provisioning task or supports the user in performing such a task. GSAC consists of a number of tools integrated via a blackboard mechanism [11], together with a control tool named the Problem Solving Control (PSC) tool (the Task Orderer in GSAC).
Figure 4.2: GSAC DSS Application Architecture.

Tools are combined and controlled explicitly using an architecture that is general enough to support a variety of control mechanisms. The concepts of control passing and data passing are separated in the architecture, and explicit control can be achieved by employing a dedicated Problem Solving Control (PSC) tool (optional) in the tool definition.

4.1.1 Blackboard Mechanism

The blackboard mechanism is analogous to a group of human experts working together and using a blackboard to write their results. When an expert sees a problem that he or she can solve, they request a piece of chalk and, having been given it, write their solution on the blackboard for other experts to see.

In its software implementation, the blackboard provides structured shared memory for problems and solutions, and provides a simple and flexible mechanism to pass data. Each tool performs a specific task which could progress a problem towards resolution. Tools examine problems on the blackboard and notify a control tool (the Task Orderer in GSAC) if they can contribute. The control tool then decides which tool will work on each problem. Tools can compete with each other (i.e. perform the same task but adopt a different approach to solving it) or co-operate with each other (i.e. work in series to resolve a problem). The user interacts with the individual tools to solve problems.

The blackboard approach has two key advantages for GSAC:

4.2 GSAC Tools

The GSAC DSS introduces a number of tools to support the user in resource assignment (see figure 4.3). Note, GSAC provides a number of tools to facilitate problem representation and exploration i.e. to support the computation and the decisions faced by the user in resource assignment. For example, tools are provided in the areas of access network resource assignment, transit network optimisation and value-added service resource assignment.

Three example tools are described in this section, each illustrating different aspects of the GSAC DSS.

  • The Task Orderer is an example of how the Problem Solving Control (PSC) tool can provide a flexible and adaptive control mechanism for decision support. The objective being to take a co-operative problem solving approach where both user and computer collaborate.
  • The MAN Access Network Generator shows how the user is supported in making a decision based upon multiple and qualitative criteria.
  • The Transit Network Configurator illustrates how Operations Research (OR) and Knowledge-Based System (KBS) techniques can be combined to provide the benefits of both techniques, while supporting a user who has no knowledge of either field.
    •

    4.2.1 The Task Orderer

    The Task Orderer tool shows how Problem Solving Control (PSC) can provide a flexible and adaptive control mechanism for decision support. The Task Orderer provides decision support to the user by continually monitoring the progress towards problem solution, advising the user of which tasks could be carried out next and recommending which of those tasks should be addressed. The user can then select which task(s) to perform by either taking or ignoring the Task Orderer's recommendations. [10] defines the term active decision support to describe systems that provide autonomous support to a user. The Task Orderer falls into this category.

    The Task Orderer has been implemented by extending the agenda-based control mechanism adopted in most blackboard systems. In a traditional blackboard system [11], whenever a tool can perform some work it informs a blackboard controller which obeys some predefined strategy in determining which of the tools that can and should be used. In the GSAC DSS this has been extended with more emphasis on user involvement. So, when tools can contribute to the problem they inform the Task Orderer which in turn, supports the user in deciding which tool to perform next. The Task Orderer monitors the reports coming from each tool and determines when they can contribute to the problem solution, presents the user with a prioritised list of tools which can be activated, and the user selects the appropriate tool.

    Producing a solution (a resource assignment) requires performing a number of steps and decisions, where the order in which those steps are performed can effect both the quality of the final solution and the efficiency of reaching that solution. For instance, decisions made on the resource assignment for an access network at a customers site will restrict options elsewhere, and possibly exclude better solutions. To date, it has not been possible to construct an ordering for these steps that is best for a sufficiently high number of cases. This reflects the lack of understanding of the structure or complexity of the problem and this precludes a fully automated approach.

    As a result it is necessary to produce a flexible decision control system that can adapt ordering to suit the particular situation. This is of major importance as constraints are placed on the solution offered to the system. Human users are often able to decide which ordering of steps is best based upon experience and domain knowledge, so it is important for the user to have a major role in ordering the steps. However, the scale of the problem does mean that it can be difficult for users to keep track of what should be done and what has been done. The Task Orderer has tried to provide this support.

    4.2.2 MAN Access Network Generator

    The MAN4 Access Network Generator (MAN-ANG) shows how the user is supported in making a decision based upon multiple and qualitative criteria. The MAN ANG tool helps to produce a set of ranked access network configurations that meet the customer's requirements. Each 'configuration' represents a possible assignment of network resources (e.g., what lines and equipment are needed to connect to the customer site, where the site will connect to the public network, the bandwidth required, etc.) This task which the tool aims to support, is traditionally undertaken as a pen and paper exercise and relies heavily upon the knowledge, expertise of an individual to determine an optimal solution.

    The basic operation of the tool is to first generate the technically feasible configurations and then to evaluate which one best meets both the customer's and the provider's requirements. The tool takes a customer's set of technical requirements and

  • determines the feasibility of the requirements for the network technology,
  • determines the capacity required5,
  • generates all feasible (given local situation) access configurations for the customer's access network and
  • establishes the best (optimal) solution that meets both the customer's and the provider's requirements.
    •  In this tool, steps (1 to 3) are based on KBS, heuristic, rule-based reasoning6 [13] allowing the Decision Support System (DSS) to deal with subject matter of realistic complexity that normally requires a considerable amount of human expertise. These steps produce a set of configurations (called a solution set) that each satisfy the customer's requirements. It is then necessary, to establish the best (optimal) solution from this set. Step 4 is based on Multi-Criteria Decision Making techniques [14], and supports the user in establishing the best (optimal) solution from the configuration solution set and assists the user in understanding the trade-offs that exist between possible solutions. For example, it may be that the most reliable solution is the least extensible solution so the user will need to consider how important these factors are.

    The mechanism behind the technique examines the performance of every possible solution against the set of desired criteria. (This optimal solution is understood to be the best solution from a service provider's point of view.) In fact, the tool provides a ranking of the solution set, based on the set of desired criteria. One obvious criterion for example, is cost but in a realistic scenario several factors may influence the overall optimal solution, for instance, installation time, the degree of reliability, the ability of the solution to evolve, etc. [15]. The tool also allows the service provider (the user) to 'weigh' the importance of each criterion. This facilitates the service provider in finding solutions that are for example, most importantly, cheap and least importantly, secure.

    The tool uses a fuzzy weighted averaging approach to Multi-Criteria Decision Making (MCDM) [16, 17]. Each configuration is assigned a fuzzy value for each of the criteria under consideration. The importance given by the user to each criterion provides fuzzy weights that are used to produce the weighted average of importance for each configuration. The weighted averages approach supports decision making in the context of multiple criteria evaluation. Extending this technique to use fuzzy set theory, to incorporate the modelling of imprecise natural language expressions (or linguistic terms, for example high, medium, low, etc.) has been very effective in hiding unnecessary details (for example, numerical data) from the user. Linguistic terms are useful here to firstly, quantify criteria that can't be easily quantified with a numerical value (e.g., security), and secondly to help provide a more user friendly presentation of numerically quantifiable criteria.

    An example is illustrated in Figure 4.4. The top window contains a number of criteria which are the relevant factors used to judge a configuration. It provides a set of sliders, one for each criterion. The user moves the sliders to specify the importance of the criteria for this solution. The phrase displayed on the right is changed in accordance with slider's position and provides a linguistic interpretation of the importance. So in the example shown in Figure 4.4, the user has stated that minimising connectivity costs is of high importance whilst indicating that a configuration with a short installation time is of low importance. When the 'apply' button is pressed the weights specified by the current position of the importance sliders are used as the basis for calculating an aggregate rating for each alternative. The aggregate ratings are displayed in the second window, ordered from best to worst. At the same time the ‘best’ alternative is displayed in more detail in the bottom section of this window. The user can experiment with different importance weightings for the criteria and hence gain understanding on how and when trade-offs occur.

    Addressing decision problems involving multiple criteria, MCDM is crucial for the decision support paradigm. [18] describes the wide applicability of Multi-Criteria DSSs and indicates their importance within the DSS research field (see also [19] and [20].) The need for the user to be involved with this class of problem is emphasised, since multiple criteria problems cannot be completely solved for a single solution independently from the user. The use of graphic facilities during the process is quite useful and necessary since humans can judge and absorb visual information much faster. When such an interactive method is embedded in a heuristic or rule-based approach the decision maker is able to change some parameters existing in the knowledge base (e.g. weights of criteria introduced in some rules). New solutions can then be generated and again judged and analysed until the decision maker considers that his/her aspirations are attained. In this way the decision maker progressively explores the feasible solutions in search of an ideal answer.

    One of the strengths of this method is that it avoids the common difficulty which decision makers have in being able to express their preferences before candidate solutions are inspected. As possible solutions are viewed the decision maker becomes clearer about the characteristics of the ideal solution and can steer the system to search in the most promising regions of the solution space. It is most important to emphasis that providing methods of dealing effectively with multiple criteria decision making is a crucial element for powerful DSSs.

    Figure 4.4: Configuration evaluation.

    The tool uses a fuzzy weighted averaging approach to Multi-Criteria Decision Making (MCDM). The decision maker expresses his/her preferences about the importance of each criterion by graphical support.

    4.2.3 Transit Network Configurator

    The Transit Network Configurator (TNC) illustrates how Operations Research (OR) and Knowledge-Based System (KBS) techniques can be combined to provide the benefits of both techniques, while supporting a user who has no knowledge of either field. The TNC supports the user in placing a set of specifications corresponding to customers' requirements on the bandwidth available across a transit network. In DESSERT, the transit network is viewed as a collection of large bandwidth circuits each with defined cost and quality and each capable of supporting a number of customer requirements. The user and the tool configure the network by placing the customer requirements on these circuits. The specification of a customer's requirements for the transit network will describe the required bandwidth, the required Quality of Service (e.g. security, reliability and availability) and usage information describing how much network traffic the service is expected to produce. The objective is to meet these requirements to a level acceptable to the customer while meeting the business objectives of the service provider (e.g. minimising costs).

    Once a solution has been produced the tool will critique the solution judging how well it meets the input criteria and reporting its opinion to the user (shown in Figure 4.5). Essentially this is achieved by comparing the original Quality of Service (QoS) requirements input to the tool with the QoS presented by the solution. Any differences are reported to the user. If the user is dissatisfied, they can alter the matching accuracy that are attached to each criteria and produce a different solution.

    Figure 4.5: The Transit Network Configurator (TNC) tool.
    An evaluation of the result produced is displayed to the user. Factors considered include utilisation, fragmentation, spare capacity, cost per unit of bandwidth and QoS provided.

    An OR technique, a Linear Programming Solver (LP-Solver), is used in the TNC [21]. It packs channels (i.e. transit requirements) into pipes (i.e. available transit connections) while optimising cost. This part of the TNC is termed the pipe filler. The pipe filler requires support functionality which pre-processes the input data into a format suitable for the LP-Solver, and interprets the output suitable for use by the rest of the TNC. The response time of the LP-Solver is proportional to the number of Quality of Service (QoS) criteria and the number of possible values of those criteria. A small change can lead to a large change in the time needed for a solution. The pipe filler estimates how long a solution will take. If it is above a threshold value it will approximate the problem to reduce the computation time. For example, if five levels of security are available initially this could be reduced to two levels with subsequent reduction in solution time and displayed to the user.

    OR techniques are used here to ensure bandwidth requirements are met while minimising a single cost function such as price. However, the best configuration may be the one that is cheapest, most reliable and best meets the usage requirements. KBS techniques are good at handling such ill-defined concepts and at producing solutions that are acceptable, though perhaps not optimal, against a number of criteria. The TNC tool contains a model that describes the type of problem that each technique is able to solve best. This is used to choose between the OR and KBS approach. The user is asked to set some parameters to help characterise the problem. These parameters are related to domain aspects and do not require any knowledge of the OR and KBS techniques. Based upon the parameters and the model the appropriate technique will be used.

    The TNC has shown how OR may be beneficially combined with simple KBS heuristics to speed computation in cases where the standard LP input tableau becomes large, i.e. when there are a large number of decision variables and constraints.

    5. Conclusion

    The research presented in this paper has illustrated, through its DSS prototype, how the application of decision support technology is a key element in making the provisioning of services (in particular resource assignment) more efficient. An important characteristic of the GSAC approach is the combination of KBS and OR techniques being effectively applied in the development of such DSSs in the field of telecommunications management. The explosive growth in the field of telecommunications has created an important opportunity for contributions from decision support technology.

    From our experience the following are key features to a successful DSS:

  • Using a flexible and evolvable approach (e.g. based on a blackboard model) enables the inter-working of a wide variety of well understood and newer state of the art techniques within a single unifying framework.
  • OR can be beneficially combined with simple KBS heuristics to speed computation in cases where there are a large number of decision variables and constraints.
  • Methods of dealing effectively with multiple criteria decision making are crucial for powerful DSSs.
  • Using graphic facilities and providing multiple views on to the problem is necessary for decision support since humans can judge and absorb visual information much faster.
    •  The results of this paper illustrate and reinforce the argument that applications of advanced computer techniques are vital support tools in this area.

    Acknowledgements

    We would like to acknowledge the support of the various people associated with the DESSERT project, all of whom have contributed to the project's results. Special thanks are due to Declan O'Sullivan of Broadcom and Tim Brown of Queen Mary and Westfield College. DESSERT is a RACE II project, partly funded by the Commission of the European Communities, DG XIII, Telecommunications, Information Industries and Innovation. The DESSERT consortium includes Broadcom Éireann Research Ltd., BT plc, Framentec-Cognitech, SEMA Group Telecom, INFORM Gmbh, Trinity College Dublin, PTT Netherlands, Queen Mary and Westfield College and Dublin City University.

    REFERENCES

    [1] Basseil R. and Gotz B.; 'Integrated Service Management for Intelligent Networks', XII International Switching Symposium, Stockholm (1990).
    [2] Minoli, D.; Enterprise Networking: Fractional T1 to SONET, Frame Relay to BISDN, Artech House, 1993.
    [3] Davison R., O'Brien P.; 'Service Provisioning in a Multi-Provider Environment', In Proceedings of the 1994 Intelligence in Services and Networks Conference (IS&N'94), Springer Verlag, 1994.
    [4] Tattersall, C., Braux, H., et al.; 'Helping Consultants Helping Customers: Decision Support for Requirements Engineering during Service Provisioning', DESSERT RACE II (R2021) Deliverable 25, 1994.
    [5] Porté, N., Valette, C.; 'A Decision Support System for Scheduling and Network Configuration Installation', DESSERT (R2021) Deliverable 25, 1994.
    [6] O'Sullivan S., et. al.; 'A Reuse Methodology for Service Management', 7th International Conference of Systems Research, Informatics and Cybernetics, Baden, Germany, 1994.
    [7] Keen P., Scott-Morton M.; 'DSS: An Organisational Perspective', Addison Wesley, (1978).
    [8] Gavish, B. (Ed.); Operations Research in Telecommunications, Annals of OR, Baltzer Scientific Publishing, Basel, 1992.
    [9] Williamson G., et al; 'Using KBS in Telecommunications', ESPRIT 1987; Achievements and Impacts Part I, North Holland, 1987.
    [10] Raghavan, S. A., JANUS; 'A paradigm for active decision support', Decision Support Systems 7, pp 379-395, 1991.
    [11] Nii. H. P.; Blackboard systems (Part I); 'The blackboard model of problem solving and the evolution of blackboard architectures' The AI Magazine, pp 38–53, Summer 1986.
    [12] Bazaraa, M. S., Jarvis, J. J. and Sherali, H. D.; Linear Programming and Network Flows, Second Edition, Wiley, 1990.
    [13] McDermott, J.; R1: a rule-based configurer of computer systems, Artificial Intelligence 19, pp 39–88, 1982.
    [14] Zeleny, M.; Multiple Criteria Decision Making, MacGraw Hill, Columbia University, 1982.
    [15] Minoli, D.; Broadband Network Analysis and Design, Artech House, 1993.
    [16] Brown T.; 'A New Fuzzy Weighted Average Algorithm', Proceedings of the IMACS International Workshop on Qualitative Reasoning and Decision Technologies QUARDET '93, (Eds) N. Piera, M. G. Singh, CIMNE Barcelona, 1993.
    [17] Zadeh, L. A.; 'Fuzzy Sets', Information and Control, No. 8, pp 330-353, 1965.
    [18] Eom, H. B.; 'The Current State of Multiple Criteria Decision Support Systems', Human Systems Management, 8, pp 113-119, 1989.
    [19] Keen, P. G. W.; 'Decision Support Systems: The Next Decade', Decision Support Systems, The International Journal, 3, 3, pp 253-265, 1987.
    [20] Sprague, R. H., Carlson E. D.; Building Effective Decision Support Systems, Prentice Hall, 1982.
    [21] Rissland, E. L. and Shank, D. B.; 'Combining cased-based reasoning and rule-based reasoning: A heuristic approach', Proceedings of 11th Joint Conference on Artificial Intelligence, San Mateo, CA, Morgan Kaufmann Pub., Inc., 1989.
     

    1DESSERT is a RACE II project (R2021), partly funded by the Commission of the European Communities, DG XIII, Telecommunications, Information Industries and Innovation.
     2 as a result of the strict and quantitative character of optimisation (OR) algorithms.
    3 Although the GSAC DSS can provide support as will be described later, only the user can decide how important the criteria are and make the final choice between resource assignments.
    4 Metropolitan Area Network.
    5 This task is further constrained in meeting a set of usage (profiles) parameters and QoS parameters.
    6 and more recently Case-Based Reasoning [12].