F. Hauck, U. Becker, M. Geier, E. Meier, U. Rastofer and M. Steckmeier
Friedrich-Alexander University of Erlangen-Nürnberg, Germany

Download paper in postscript (103k) or gzip compressed postscript (28k).

The CORBA architecture defines the semantics of the interaction of distributed objects. These semantics are fixed and not extendable. CORBA services can extend the basic functionality of an ORB, but they are based on these fixed semantics.

AspectIX is an open and more flexible architecture than CORBA, but an AspectIX implementation can also host CORBA-compliant applications. AspectIX adopts a fragmented object model, which means that each client owns a local part of the distributed object and that these local parts (called fragments) can interact with each other. A local fragment can be intelligent and carry a part of the distributed object's functionality, or it can be dumb and act as a stub only, as in the CORBA-compliant AspectIX profile.

A client can configure different nonfunctional aspects of the interaction semantics using a generic configuration interface. A local fragment implementation may be transparently replaced by another one if it cannot fulfill the requirements of a new configuration. We will present examples how to use the configuration interface for problems in wide-area systems (replication, consistency), mobile agents (mobility), and process control systems (real-time constraints, fault tolerance).

Appears: Work-in-progress Presentations, Wednesday PM

T. Kramp and R. Koster
University of Kaiserslautern, Germany

Download paper in postscript (335k) or gzip compressed postscript (71k).

The demand for QoS-support by middleware platforms is continuously increasing, particularly in application domains such telecommunications or multimedia. Current object-oriented middleware, however, is not well suited to fulfil this demand. In this paper, we propose a service-centred approach to QoS-supporting middleware, introducing the notion of smart proxies to achieve location transparency of services, not only objects.

Appears: Work-in-progress Presentations, Wednesday PM

P. McDaniel, P. Honeyman and A. Prakash
University of Michigan, USA

Download paper in postscript (4307k) or gzip compressed postscript (125k).

An advantage of today's high speed networks is the ability to support group communication. Applications that support group communication allow the free exchange of ideas and data in real time, regardless of the physical distance between the participants. Unfortunately, support for additional protocol features such as reliability, secrecy, and total ordering in the multicast context requires more bandwidth and greater complexity than in traditional point-to-point communication. In this paper we describe a middleware software layer and associated API that attempts to minimize these requirements by providing multiple secure channels based on IP multicast within the same logical group. Named LGSC (lightweight secure group communication), the software provides the important features needed by a group application: reliable delivery, best-effort delivery, and security. In providing both reliable and unreliable channels, an application need pay only for the delivery assurances it needs. We conclude with a description of our implementation and supporting performance data.

Appears: Work-in-progress Presentations, Wednesday PM

P. Murray, R. Fleming, P. Harry and P. Vickers
Hewlett Packard Labs, UK

Download paper in postscript (529k) or gzip compressed postscript (149k).

Somersault is a platform for developing distributed fault-tolerant software components and integrating these critical components with other components into distributed system solutions. Critical application processes are mirrored across a network, with each critical process being replicated in a primary and secondary. Replication of processes and recovery from the failure of a replica are handled transparently. Somersault provides a fault-tolerant communication transport protocol, which can be plugged into an Object Requent Broker, a combination which achieves replication transparency.

Somersault has been developed at Hewlett-Packard Laboratories, and runs on standard operating systems such as Unix and NT. Somersault has been through two iterations of algorithm design, implementation and optimisation in order to achieve high message throughput and fast recovery from failure. At the time of writing it is undergoing industrial trials and detailed performance testing.

Appears: Work-in-progress Presentations, Wednesday PM

C. Peter and F. Puntigam
TU Vienna, Austria

Download paper in postscript (392k) or gzip compressed postscript (75k).

Coordination is one of the most important tasks in concurrent, distributed program development. Conventional interface definitions like the CORBA_IDL do not support the programmer concerning coordination. Messages sent at the wrong time may cause errors to occur at remote sites, where they are difficult to handle. Much design effort is required to avoid them. Process types provide means to ensure that all clients are coordinated before sending a message to a server so that no "message-not-understood-errors" can occur even if the understandability of messages depends on the server's state. An extension of the CORBA-IDL allows us to combine the benefits of process types with the basic object design of CORBA. Type errors caused by wrong message orderings are detected at run time beore these messages are sent to a remote client; the exceptions can be handled locally. A subtyping relation allows us to inherit restrictions on message orderings.

Appears: Work-in-progress Presentations, Wednesday PM

T. Preuß, J.-H. Syrbe and H. König
Brandenburg University of Technology at Cottbus, Germany

Download paper in postscript (938k) or gzip compressed postscript (339k).

Open Service Markets (OSM) with a multitude of service types, service providers and service customers have come into being. Certain services are offered by different service providers at the same time, but differing in their service quality. A core issue for OSMs is the question how service providers and service customers come together and how the interaction between them takes place. Most approaches are based on the trading concept and are dedicated for a single service use only. Based on an analysis of the OSM model we sketch the Service Supplier Relation (SSR) model supporting the obligatory provision of repetitively used services. We have drawn inspiration from sub-contracting industry which is widely accepted e.g. in the automotive industry. After discussing the SSR model we suggest strategies for the implementation of Service Supplier Relations.

Appears: Work-in-progress Presentations, Wednesday PM

P.-G. Raverdy and R. Lea
Sony Computer Science Lab and Sony Distributed Systems Lab, Japan

Download paper in postscript (845k) or gzip compressed postscript (108k).

The DART (Distributed Adaptive Run-Time) project is developing an adaptive software environment for general purpose distributed applications. The goal of the project is to provide a software run-time that will allow application authors to quickly develop distributed software, such as network or web software, without having to deal with the details of the distribution technology. A key feature of the run time is its ability to reconfigure, or adapt itself, to both application requirements and to the system and networking environment, To achieve this adaptation we use meta-level object programming techniques and run time monitoring.

Appears: Work-in-progress Presentations, Wednesday PM

L. Sacks, M. Verdier, O. Prnjat, M. Loryman, M. Wittig, M. Kande, B. Bushan, S. Mazaher, D. Maillot, C. Autant and N. Ganivet
University College London, UK / GMD Fokus, Germany / Norwegian Computing Center, Norway / Telis, France / Alcatel ISR, France

Download paper in postscript (727k) or gzip compressed postscript (94k).

This paper describes the TRUMPET service management architecture. This architecture was developed to support the aims of the TRUMPET (Inter-Domain Management with Integrity) project: to investigate secure, high integrity interactions between administratively separate bodies concerned with the provisioning of broad-band telecommunications services. To allow these issues to be fully investigated a system was designed that included a number of bodies, or organisations, and in which these players interact over a mixture of technologies - specifically Java, CMIP and CORBA. These technologies were deployed so as to provide a technology independent interface information model following the TMN recommendations. The service architecture provides ATM Virtual Path connections, within specified Quality of Service parameters, across two or more Public Network Operators offering service to a Values Added Service Provider; who in turn is offering the full end-to-end or 'One Stop Shopping' service to a number of customers.

Appears: Work-in-progress Presentations, Wednesday PM

I. Welch and R. Stroud
University of Newcastle-upon-Tyne, UK

Download paper in postscript (839k) or gzip compressed postscript (238k).

Distributed languages such as Java enable applications to be built out of third-party components that are downloaded and integrated into running applications as required. This has great advantages in terms of productivity, and also eases the problems of software distribution. However, applications may require such components to be modified in order to ensure fault-tolerance or security properties. Without source code, changing the functionality of components is problematic. Also, if components are transient, then the changes would need to be reapplied every time the components are downloaded and linked into the application. In this paper, we present a solution implemented in Java (JDK1.2) based on a metaobject protocol approach. This relective technique allows dynamic reconfiguration of the behaviour of downloaded components with no need for access to source code and only minor modifications to the applications that make use of the components. It allows the underlying mechanisms that implement the behaviour of components to be adjusted locally and incrementally without exposing unnecessary implementation details. As an illustation, the paper demonstrates how access control methods can be added transparently to selected component operations.

Appears: Work-in-progress Presentations, Wednesday PM