Resource federation in grid using automated intelligent agent negotiation

Abstract

This paper provides a novel approach for enabling grid users to perform resource federation using intelligent agent negotiation. The features of agents including autonomous and distributed are designed to address the federation problems in grid, such as resource selection and policy reconciliation. Early work of resource federation is still highly depending on resource administrator (a person who manages the resources for corresponding institute). In this paper, automated agent negotiation in grid is presented to maximize the satisfaction level of federation outcome for involving parties. Using a new conflict resolution technique–creative negotiation, is introduced in solving communication problems during negotiation. A framework of automated negotiation with conflict resolution–select, match, negotiate and expand (SMNE) protocol, is proposed to guide resource administrators in performing resource federation with intelligent agent. According to the experimental result, the proposed SMNE protocol is able to guide resource administrator in conflict resolution when negotiated parties are unable to make further a concession in resource federation.

Introduction

Grid computing [1] is an infrastructure to enable heterogeneous resources to be shared dynamically between various parties. The parties involved can federate for a period of time as a team or given a name as Virtual Organization (VO) [2], [3]. The process of aggregating resources within distributed VOs is called resource federation. The federation of resources is not limited to computer and storage, but also other scientific instruments and devices. The accessibility of each resource has a set of rules imposed by corresponding owners to avoid overloading of tasks during resource federation. This set of rules is known as policy in this paper.

A policy is a plan of actions to guide decisions and actions of individuals or groups. The term “policy” mentioned in this paper is referring to resource usage policy [2] which defines the access and usage over resources. As shown in Fig. 1, a policy is detailed with different criteria. Those criteria listed out the terms and conditions in resource accessibility, such as the object to be shared (data server), the parties who are allowed to utilize (UTAR members) and restrictions of sharing (share data storage for 3 days with maximum storage of 2 TB).

Before resource federation, every local administrator may impose a set of resource usage policy [4] on a particular resource (e.g. CPU usage, disk storage and network bandwidth) to allow a third party (a party who responsible to construct a VO) to perform resource selection [5], [6]. The third party or VO administrator will classify and select the appropriate resources from different domains to construct a VO by referring to resource usage policies. The VO administrator is responsible in the selection of the best candidate to join a VO. However, the resource selection is a difficult task since every local administrator has different interests and aims (self-interest) when contributing resources to a VO. In most of the cases, policy applied on a particular resource is complex, thus increasing the burden of a VO administrator. Sometimes, much effort and time are spent to generate a mutual acceptance between VO administrator and participant but failed. Would it be a better solution to look for another VO participant (may time and effort consuming) or resolve the current negotiation problem? What if the current VO participant is the only available candidate?

Due to these difficulties, an automated policy reconciliation method is needed to reduce the workload for VO administrator in resource federation. In this paper, a concept of Creative Negotiation [7] which yet to be applied in automated conflict resolution with agent technologies is proposed. Using autonomous agent in resource federation becomes state-of-the-art in current grid domain [8]. However, the feasibility of autonomous agent with appropriate features is still an opened issue. In this paper, several policy reconciliation techniques from manual to automatic are being studied, and a novel automated negotiation and conflict resolution protocol is proposed using intelligent agent for resource federation. A comparison of different approaches in resource federation is also analyzed via a multi-agent system (MAS) grid simulator.

The rest of the paper is as follows; Section 2 discusses the related work on agent negotiation in resource federation. Section 3 gives description on creative negotiation, policy representation and SMNE protocol. Two scenarios require creative negotiation are also described in this section. Section 4 presents the implementation of MAS in grid simulation. Section 5 explains the experimental setting, evaluation criteria and experimental results. Last section is the conclusion with future work.