An Automated Negotiation Based Approach for QoS-aware Service Provision and Service Composition

Abstract

Service-centric computing has recently become an important paradigm for corporate IT resource management. While software applications are traditionally specified and implemented as a set of functions that are uniform to all users, the new paradigm of service-centric computing is characterized by the fact that the same software service can be delivered with a different price, response time, availability, and other non-functional attributes to accommodate different modes of use. These non-functional attributes together are referred to as Quality of Service (QoS). Given that the provision of software service relies on dynamic environmental conditions, such as network traffic and the demand and supply on the service market, it is flexible and desirable that the user negotiates with service providers dynamically to reach agreements on the QoS of the service to be provided.
If the service that the user requires can be fulfilled individually by several providers with the same functionality but with different QoS, the user (or broker who represents the user) can negotiate with all providers simultaneously to reach an agreement on the QoS with one provider, which is referred to as one-to-many concurrent negotiation model for a single service provision. It can be considered as several one-to-one individual negotiations that are carried out concurrently, each of which involves the user and one provider. A significant issue in this model is the coordination of all negotiation threads to generate offers by making concessions. Most existing studies on one-to-many concurrent negotiation adopt the concession function from one-to-one bilateral negotiation, which determines the utility level the user is going to concede to for each provider individually. Without coordination across different negotiation threads, the user may concede too much in negotiating with one provider to seek an agreement even though a good chance exists that a higher utility can be obtained with another provider than with the existing provider. Thus, adjusting the concessions on the basis of negotiations with other opponents in the context of one-to-many negotiation becomes impossible. To address the limitations of such independent concessions, we propose a novel concession strategy that coordinates multiple concurrent negotiations with different providers. We show that the problem of determining the concession with each provider can be modeled as a multi-objective optimization problem. We are able to identify some important properties of the optimal solution, which enables us to obtain the solution in an elegant manner. We further conduct simulation experiments to demonstrate the effectiveness of our proposed approach.
When a single service provider cannot meet a user’s requirements, the user needs to compose different services with different functions to meet his requirements and this is referred to as service composition. When creating a new composite service, negotiation allows a service provider to offer the service with QoS properties customized to the needs of a user. Negotiation is particularly useful when a feasible solution is not readily available on the service market. When negotiation is used for service composition, the user (or the broker) should negotiate with providers who offer different functions and reach an agreement with each individual service provider on the QoS attributes such that the agreed QoS together can satisfy the user’s requirements (QoS constraints) on the composite service. Given that bilateral automated negotiation tactics requires the specification of the reservation value for each QoS attribute, an important challenge in negotiation based service composition is to determine the reservation values for each service based on the user’s requirements on the composite service.
We first present a base approach that sets the reservation values proportionally. When the user has information about the preferences of different providers on different QoS attributes, the user can make trade-offs across different QoS attributes among different providers to set the reservation values and increase the likelihood of success of the negotiation. We then presented an enhanced approach that extends the base approach by incorporating the preferences of providers to determine reservation values. The problem of determining a user’s reservation values for component services can be modeled as a multi-objective optimization problem; this problem can then be transformed to a single-objective optimization problem by using a max-min approach. The max-min transformation may lead to weak Pareto optimal solutions and we show how to improve weak Pareto optimal solutions into strong Pareto optimal solutions. If some providers accept offers before the negotiation process deadline, we show how to dynamically adjust the reservation values for other ongoing negotiations. Simulation experiments have been conducted to study the performance of the base approach and enhanced approach.

Date of Award	5 Feb 2016
Original language	English
Awarding Institution	City University of Hong Kong
Supervisor	J Leon ZHAO (Supervisor) & Xiaoyun Sherry SUN (Supervisor)