The improved power consumption laxity-based (IPCLB) algorithm is discussed to select one of servers so that the total power consumption of a cluster can be reduced. However, a load balancer has to collect a state of every current process on servers of a cluster to calculate the estimated power consumption of each server. In addition, it is difficult to precisely estimate the power consumption of each server since the state of each process on the server is changed during the estimation. Especially, a process might terminate before the termination time is estimated if the computation time of the process is shorter than the communication delay time between the load balancer and the server. In this paper, we assume the computation time of each process is shorter than the communication delay time. Then, we propose a delay time-based (DTB) algorithm to select a server for each request process so that the total power consumption of a cluster to perform processes on the server can be reduced. In the DTB algorithm, it is not necessary to collect a state of every process on each server to estimate the power consumption laxity. In addition, the minimum computation time of a process is not required to be a priori defined in the DTB algorithm.

In peer-to-peer (P2P) applications, a group of multiple peer processes (peers) are required to cooperate with each other. In this paper, we discuss a heterogeneous hybridtime group communication (HHT) protocol which takes advantage of the linear time (LT) and physical time (PT) to causally order messages in a scalable heterogeneous group. It depends on accuracy of each physical clock and minimum delay time between a pair of peers how messages can be ordered. In this paper, we consider a heterogeneous type of group where the clock accuracy of each peer and the minimum delay time between every pair of peers are not the same. In group protocols, even if a pair of messages are ordered in the protocol, the messages may not be causally ordered. Thus, some messages are unnecessarily ordered in the protocols. In this paper, we show the number of messages to be unnecessarily ordered can be reduced in the HHT protocol. In a scalable group, it is not easy, maybe impossible for each peer to hold information on the clock accuracy and minimum delay time of every peer. In this paper, we newly consider a two-layered model of a heterogeneous group to reduce the information which each peer has to hold.

A transaction is assigned with a purpose which is a collection of roles. Suppose a transaction T1writes an object o2after reading an object o1and then another transaction T2reads the object o2and writes an object o3. Here, data in the object o1might flow into o3via o2. Unless T2is granted a read access right of the object o1, illegal information flow occur. In order to prevent the illegal information flow, T1marks the object o2with the purpose of T1. T2cannot read o2unless the purpose of T2includes a read right of o1. In result, the throughput is degraded. Objects whose information may flow into an object o are source objects of o. An object is timed out if it takes some time units after the object is lastly written. While there occur no illegal information flow in our purpose marking (PM)protocol, transactions which imply illegal information flow are aborted. We evaluate the PM protocol in terms of how many transactions are aborted.

The paper presents a model for the dynamic integration of global ontologies from peer ontologies in peer to peer data and knowledge sharing (P2P/DAKS) networks, and demonstrates its effectiveness. The peer ontologies from peer schemas (possibly heterogeneous) have to be created within some rules, as discussed. The integration is carried out with the help of a graph based inclusion algorithm to determine the semantic closeness of inter-ontology concepts. The effectiveness of the model is demonstrated by comparing its results with those of a WordNet-based approach.

Service supported by an object is modeled to be a set of methods and quality of service (QoS). In addition to finding a peer which holds a target object, it is critical to discuss what peer is allowed to manipulate the target object in what method. In this paper, we take an acquaintance approach to finding and manipulating objects in P2P overlay networks. An acquaintance peer of a peer p is a peer whose service the peer p know and with which the peer p can directly communicate. If an access request is issued, a peer has to find where a target object exists, how to obtain an access right on the target object, and how to manipulate the target object. In this paper, we discuss ways to obtain results which satisfying an access requests by the cooperation of acquaintances. Acquaintance peers of a peer p may knows different information on target peers since it takes time to propagate change information of the target peers and peers may be faulty. Here, it is critical to discuss how much a peer can trust each acquaintance. We define the trustworthiness of an acquaintance peer in terms of the acquaintance relations among the peers.

We present here an approach for dynamic ontology integration for a multi-agent environment, in which each agent holds the ontologies of its acquaintances (i.e., other agents of its interest) as the integrated partial global ontology, which is essential to interpret the local schemas for inter-agent operations. This integration has to be carried out whenever a new acquaintance is added or when the local ontology of an acquaintance changes. The approach described is general (i.e., independent of any particular thesaurus) and it carries out the integration automatically except for minimal unavoidable human inputs to resolve semantic conflicts if discovered in the process.

An object is a unit of resource distributed in a peer-to-peer (P2P) overlay network. Service supported by an object is modeled to be a set of methods. It is critical to discuss which peer can manipulate an object in which method, i.e. only a peer granted an access right (permission) can manipulate an object. First, an application has to find target peers which can manipulate a target object. In this paper, we take an acquaintance approach to finding and manipulating objects. Here, it is critical to discuss how much a peer trusts each acquaintance peer. We define the trustworthiness of each acquaintance peer in terms of how satisfiable a response from the peer is for each access request.

The paper presents a general architecture for a P2P data sharing facility within a multi-agent framework, where peers as autonomous high-level nodal agents cooperate with each other to solve global tasks. A node may have several lower level local agents including local databases and partial global ontologies. In addition there are also minder agents coordinating the activities of the peers that offer the same type of service, thus providing fault-tolerance. The ability of this architecture in data and task sharing has been demonstrated by considering query processing and directory update strategies.

A theoretical grounding is provided for a cooperating knowledge-based systems (CKBS) model which is based upon agents, cooperation blocks, and cooperation block hierarchies. Our model describes the requirements for task decomposition, negotiation, cooperation and coordination, fault tolerance, and recoverability, these requirements in turn defining a holonic system. The behavioral properties of our model are described using state transition diagrams and properties of correctness and termination are proven.

Abstract This paper presents a model of fault‐tolerant holonic manufacturing systems (HMS) where each holon's activities are controlled by an intelligent software agent. Multiple agents schedule actions, resolve conflicts and manage information to produce, transport, assemble, inspect and store customized products. Our model provides robustness and distribution transparency across a shop‐floor where unpredictable failures occur with machines, control software and communication networks. Each autonomous holon is composed of a hierarchy of large‐grain functional components where interaction is carried out by user‐defined cooperation strategies. These strategies enable holons to coordinate their behaviour through exchanging messages and sensing/actuating of their shared environment. Therefore, holonic agents can select suitable rescheduling and recovery mechanisms to tolerate faults and keep the manufacturing system working. We also propose how the IEC 1499 standard (Function Block Architecture) for distributed control systems could be used to implement our model. The model presented here is a crystallization of some abstract concepts from a generic cooperating agent system, with suitable extensions to meet the criteria of the ongoing HMS project. Copyright © 2001 John Wiley & Sons, Ltd.

This paper presents an architectural framework for cooperating knowledge based systems (CKBSs) with parallels drawn from the multiagent systems of DAI. A CKBS is distinguished from a multiagent system by its need to provide a workable approach for real-world distributed applications. The framework proposed considers only interagent activities in what is called transaction-oriented processing. The framework, based largely on well-tested computer science concepts, provides for a multilayered edifice with information transparency, and a multilevel schema to suit different user expertise. It permits dynamic definition of cooperation strategies for different tasks as required, in a high-level language providing relative ease of use. A particular novelty is the interpretation of actions as side-effects of update operations on action tuples transmitted among agents via what are called shadows. This provides the generality needed. Effectiveness, flexibility, and ease of use are some of the key considerations.

Distributed directory systems are intended to facilitate communications within the ISO/OSI framework, by holding network addresses and other relevant details of what are called communication entities, on an international scale. They can be viewed as constituting a distributed database system in a loosely couple framework. An interesting departure of directory system from the traditional DDBs is the absence of any global directory of data distribution, which often forces the directory system to use a navigational or a multicasting search strategy based on hints or partial knowledge. Such a strategy must also prevent the search from ending up in blind alleys or cyclic loops. There are also some special problems relating to issues such as aliases, synonyms, confidentially and updates. The directory system described here is based on the architecture of the PRECI* distributed database system and was developed under contract for British Telecom. The architecture has been developed to conform to emerging international standards for directory systems as developed by CCITT and ISO. (The architecture described here is based on the standard as of early 1986). Detailed functional specifications have been produced and an implementation, based on the specifications, is being planned.

PRECI* provides a generalised architecture for a decentralised distributed database system capable of supporting a heterogeneous collection of pre-existing nodes. It permits data replication, subject to nodal access controls, and a number of user facilities, including both location-transparent and location-dependent queries in a relational language. A subset of these features, dealing with distributed queries in a decentralised environment, has been implemented using two homogeneous PRECI nodes within the same computer. This system is written in C under Unix, as described in this paper.

Data integration in a distributed database refers to the production of union-compatible views for similar information expressed dissimilarly in different nodes. Such a facility is needed for location transparency and for easier formulation of global queries over the apparently incompatible data aggregated from different nodes. This paper examines the issues in data integration within a relational context, and proposes a solution based on special relational constructs which produce union-compatible relations. The advantages of this approach over others have also been discussed.

We present here a tuple placement technique, with two main options and several variations, to fulfil the differing requirements of small and large databases efficiently. The main objective is to provide fast random and sequential access by primary key, but effects on memory size, secondary indexes, addition of new domains and variable length tuples have also been investigated: and a comparison with a traditional technique given. An option of this technique has been implemented in the PRECI/H database system, and another option is intended for implementation in a small database system, particularly in connection with a heterogeneous distributed database project (PRECI*).

A research prototype of a generalised distributed database system called PRECI* is currently being developed at the University of Aberdeen in collaboration with a number of research centres, mainly in Britain. The system is fully decentralised, with both retrieval and update facilities, permitting heterogeneous and even pre-existing databases as nodes and supporting links to other (external) distributed databases. The system supports both location transparent and transaction-oriented queries to fulfil differing user requirements. Nodal autonomy, data integration, metadata and staged binding of queries may be seen as its other main features. The basic design of the system is now complete, and a partial implementation is in progress.

The hash trees method of external hashing is known to have advantages for certain types of primary key distribution. In this paper the value of the method as a general indexing technique—for secondary keys as well as primary keys—is assessed, and a comparison with the B-trees method is presented.

PRECI is based on a canonical data model potentially capable of supporting user views of other models – notably CODASYL and relational ones – through local schemas and appropriate data manipulation languages. The canonical global schema consists of normalised relations and is backed up by a storage and a data dictionary. The model is being implemented at Aberdeen University as a generalised database system, to be used primarily as a tool for research in databases, with a modular design approach so that future changes can be incorporated easily. The CODASYL and relational subschema facilities are currently being developed; a relational algebra to be used for DM commands from FORTRAN programs has been provided. The storage and access strategy in PRECI is based on internal record identifiers (or surrogates) created largely in entity-identifier order. Entity records can be accessed very quickly – both randomly and sequentially – by surrogates or entity identifiers, partly with the help of a novel indexing technique, called hash tree, which is based on data compression and hashing. The model provides maximal data independence through load-time and run-time binding. The local user may define access paths independently of the storage schema. Data items, record types and set types can be added to or deleted from the global schema; the storage schema can be restructured and data pages and indexes can be easily reorganised. Changes in one schema do not require recompilations of the unaffected schemas. The DBA has complete control over the storage and indexing strategy which he can manipulate to improve performance. Five levels of optimisation are provided to enhance execution efficiency and minimise memory usage. Certain integrity checks are also carried out during run-time.

Based on the concept of a canonical data model, a global schema is proposed for a generalised data base system capable of supporting interfaces to other systems—notably Relational, Codasyl and Adabas—through appropriate local schemas and data manipulation languages. The proposed global schema consists of entries in the form of normalised relations describing data and data relationships. The investigation of the Relational, Codasyl and Adabas interfaces demonstrates the capacity of the model to support all of them effectively. However, the model is found to impose on the local models certain constraints which help remove inconsistencies and enhance user facilities. The findings of this study are being implemented in a prototype called PRECI.