urn:uuid:1b4d9c27-a1d7-347e-914d-bd3cefd8d19a Yamine Ait Ameur
Room 701, Claremont Tower
Several web services compositions languages and standards are used to describe different applications available over the web. These languages are essentially syntactic ones and do not offer any guarantee that the described services achieve the goals they have been designed for. The objective of this talk is twofold. First, it focuses on the formal modelling, design and validation of web services compositions using the Event_B method. We suggest a refinement based method that encodes BPEL models decompositions. Second, we show that relevant properties formalized as Event_B properties can be proved. We will focus on transactional properties. At end, we discuss the basic architecture of the BPEL2B Rodin Plug-in tool encoding the proposed approach. ]]> urn:uuid:fe81a05a-7a5b-76c0-b352-f67cea46af21 2010-03-04T11:33:22+00:00 Dr Peter Popov
Room 701, Claremont Tower
The talk will present a Bayesian assessment method suitable for software system built with diverse channels (software diversity) and demonstrate its applicability on a range of case-studies: assessment of an on-demand protection systems, selection of the most suitable pair of RDBMS (SQL servers) from a pool of products available off-the-shelf, a managed on-line upgrade of a web-service. The talk will also discuss the issues of scaling the method to more complex structures and will present early results from applying this ‘multiple views’ approximation on contrived examples. The accuracy of the Bayesian predictions achievable with the method is explicitly controlled using advanced statistical methods. About the presenter. Dr Popov has been with the Centre for Software Reliability, City University London since 1997. Currently he is Reader in Systems Dependability and Director of Innovation at CSR. He holds a PhD degree in computer engineering from the Polytechnic University (KPI) in Kiev, Ukraine. His research interests include methods for software and systems dependability assessment based on probabilistic modelling; using software diversity for building reliable software systems from off-the-shelf software. More recently he became involved in developing methods and tool support for interdependency analysis of critical infrastructures. ]]> urn:uuid:280f2e6c-2f8e-561a-20cf-0399d81bc05d 2010-02-11T14:21:52+00:00 Marco Serafini
Devonshire G21/22
We are increasingly relying on computer systems to carry out critical tasks with very high degrees of dependability. For example, safety-critical systems have long been used in the aerospace domain. The trend towards storing and managing personal information online is extending the need for dependable services to a growing range of applications, from emailing, to calendars, to the storage of photos. High dependability requirements can be met with higher assurance by systems that are able to tolerate worst case failures, which are often named as Byzantine. Byzantine Fault Tolerance (BFT) is required by certification authorities for most safety-critical applications and is being advocated for critical Internet services too. BFT can be achieved by replicating logical functionalities over multiple physical nodes of a distributed system. However, the entailed replication costs are quite high. This talk points out that it is possible to reduce these costs by using optimistic fault models. In fact, Byzantine faults represent the exception rather than the rule since faults are often of more benign nature. The talk will review three modern technology trends which can be leveraged for cost reduction using optimistic fault models. The first trend is the foreseen raise in the ratio of transient faults compared to permanent faults in modern hardware due to reduced geometries and lower voltage levels. The second is the development of trusted nodes which only fail in a restricted, benign manner. The third is that replication clusters are relatively small, even in large-scale systems. The talk will then focus on the last trend and present a novel algorithm, called Scrooge, which can be used to achieve low-cost BFT in large scale or geographically-distributed systems. ]]> urn:uuid:a640ede3-5a86-e6ae-07e5-e21bbaca1278 2010-03-10T11:55:52+00:00