Dominik Dahlem

The Connected States of America: Quantifying Social Radii of Influence

Human dynamics are inextricably intertwined with the social, geographical and economic environment. The continuous flux of people communicating as well as migrating, commuting, and traveling inevitably spans acquaintances across geographic space that is far from random and exhibits regular patterns. For instance, it has been shown that the probability of being acquainted with someone is closely related to the inverse distance between them. In this paper we investigate aggregated mobile phone call detail records from a large US cellular operator and map them into space to characterize the social radius of influence at two different scales: communication and mobility. We discover that scaling properties with respect to population agglomeration are similar to those discovered for other indicators of cities. We also discover spatial community structures that are divorced from administrative boundaries, and use them to quantify the different social radii of influence discovered from the data.

Decentralising a service-oriented architecture

Service-oriented computing is becoming an increasingly popular paradigm for modelling and building distributed systems in open and heterogeneous environments. However, proposed service-oriented architectures are typically based on centralised components, such as service registries or service brokers, that introduce reliability, management, and performance issues. This paper describes an approach to fully decentralise a service-oriented architecture using a self-organising peer-to-peer network maintained by service providers and consumers. The design is based on a gradient peer-to-peer topology, which allows the system to replicate a service registry using a limited number of the most stable and best performing peers. The paper evaluates the proposed approach through extensive simulation experiments and shows that the decentralised registry and the underlying peer-to-peer infrastructure scale to a large number of peers and can successfully manage high peer churn rates.

A Service-Oriented Peer-to-Peer Architecture for a Digital Ecosystem

Service-oriented computing is becoming an increasingly popular paradigm for modelling and building distributed systems in heterogeneous, decentralised, and open environments. However, proposed service-oriented architectures are usually based on centralised components, such as service registries or service brokers, that introduce reliability, management, and performance issues. In this paper, we present a fully decentralised service-oriented architecture built on top of a self-organising peer-to-peer infrastructure. This architecture is especially designed to support digital ecosystems due to its low deployment and maintenance cost and inherently decentralised nature.

Towards Improving the Availability of Service Compositions

The proliferation of open Internet-scale service-oriented platforms based on standards, such as WSDL, SOAP and BPEL, enables the composition of independent Web services into new value-added services. Such service compositions define the information flows between autonomous and potentially heterogeneous services across the boundaries of independent provider organisations. The availability of individual services in such digital ecosystems is likely to be variable due to fluctuating usage load and resource limitations imposed by a service providers infrastructure. This problem becomes more acute as the number of services in a composition increases. This paper presents a mediation model for improving the availability of composed services. The mediation model masks failures in a service composition by transparently selecting (and executing) an alternative composition at runtime. Service consumers use a common interface to a set of functionally equivalent service compositions while a selection mechanism identifies the most suitable (alternative) service composition. An evaluation of our implementation of the proposed mediation model demonstrates that the consumer perceived availability of value-added services can be improved significantly.

Flexible application rights management in a pervasive environment

The development and wider use of wireless networks and mobile devices has led to novel pervasive computing environments which pose new problems for software rights management and enforcement on resource-constrained and occasionally connected devices. Software vendors are, however, still applying old usage rights models to a platform where application rights will be specified, managed and distributed in new and different ways. The characteristics of pervasive environments, such as occasional connectivity, require the introduction of more flexible usage rights models, such as audit-based model, that do not assume the availability of network connections. In this paper we describe a pervasive application rights management architecture for both desktop and mobile applications that provides an integrated platform for the specification, generation, delivery, and management of application usage rights based on Web services standards. We also introduce flexible usage rights models required by pervasive environments that can be embedded in target applications using aspect-oriented technology.

Waiting Time Sensitivities of Social and Random Graph Models

Large scale simulation studies, such as routing in communication systems, rely on complex networks. However, the specific network structure of the application domain is often neglected in analysing network behaviour with the consequence of limited ability to conduct a sensitivity analyses. Network evolution models have been developed to account for structural properties of social graphs, such as the Internet, and enable a model-based analysis on different scales. Despite some analytical results with respect to queueing metrics, little work has been done to frame network evolution models in the context of queuing systems. This paper addresses this issue discussing social and random network evolution models and their application to queueing systems. When the stability criterion holds, analytical solutions using Jacksons Theorem and Littles Law exist and provide valuable insights into a number of queueing performance metrics. Specifically, a sensitivity analysis of the mean total waiting time versus the utilisation is carried out which shows a significant difference between the social and random network models. Understanding the impact of the structural features of complex networks is important in the design of simulation studies and operating real world systems. This is especially true when adaptive algorithms are employed and insights of their performance compared to analytical scenarios are required.

Collaborative Function Approximation in Social Multiagent Systems

Distributed Task Assignment is a convenient abstraction for load-balancing applications, workflow systems or supply-chain management. The topological features of such task networks are far from random but instead resemble that of small-worlds and scale-free networks. The agents interaction is accordingly prescribed by this network structure. Simulating decentralised optimisation algorithms using the mathematical framework of queueing theory, it has been shown that knowledge of a neighbours queueing state facilitates the minimisation of the accrued delay in a network. Therefore benign agents that have the same neighbour can share their experience and collaborate in training the function approximator according to the SARSA(0) gradient-descent update rule. The function approximator resides on the target node and its learnt state-action value mapping is shared among all nodes connecting to it. This setting is evaluated empirically using SARSA(0) reinforcement learning with the standard

Globally Optimal Multi-agent Reinforcement Learning Parameters in Distributed Task Assignment

\epsilon

Promoting evolution by abstracting over implementation structure

-greedy policy and the weighted policy learner. We show that under certain conditions this leads to improved system performance compared to individually trained function approximators.

Workshop on Pervasive Urban Applications

Large-scale simulation studies are necessary to study the learning behaviour of individual agents and the overall system dynamics. One reason is that planning algorithms to find optimal solutions to fully observable general decentralised Markov decision problems do not admit to polynomial-time worst-case complexity bounds. Additionally, agent interaction often implies a non-stationary environment which does not lend itself to asymptotically greedy policies. Therefore, policies with a constant level of exploration are required to be able to adapt continuously. This paper casts the application domain of distributed task assignment into the formalisms of queueing theory, complex networks and decentralised Markov decision problems to analyse the impact of the momentum of a standard back-propagation neural network function approximator and the discount factor of