Research Group Meetings

Our research group meetings takes place on Thursdays, 2-3pm in Room 218 in the Huxley Building (unless stated otherwise).
If you would like to give a talk, please contact Evangelia Kalyvianaki.

Autumn 2009

Title: Evolving Landscapes of Social Networks
Speaker: Eric Yu-En Lu (University of Cambridge)
When: Thursday 3/12/2009, 14.00, Huxley Building 218

Abstract: Social networks have been shown to exert profound influence over or to correlate strongly with complex social and biological phenomena. The classical results in this area span well over 40 years and have emerged out of a wide variety of studies such as knowledge diffusion of antiobiotics, human mutations, and obesity, to name a few. These results exemplify how understanding the underlying social process can lead to explanations in the wide phenomenon. In the past decade, the proliferation of information systems and web services have enabled human data collection in an uprecedented scale. Recent advances in online social networks such as twitter, Google maps, and flickr photo streams already give sharp gauges to the underlying human perception and activity.

In this talk, I will go through results from phone, online social networks, and the primate dataset to discuss recent trends in social network formation and some ideas we are exploring.

Short Bio: Dr. Eric Yu-En Lu is a Research Associate in the Computer Laboratory, University of Cambridge and a Junior Research Fellow in Wolfson College, the most cosmopolitan college in Cambridge. His primary research interest is in complex systems including topics like high-dimensional data analysis and mining, stochastic modelling, and very large stream databases. With collaborators, he has also worked in high throughput proteomics, structural biology, and quantum physics. Prior to his PhD in Cambridge, he obtained his BA and Masters from National Taiwan University. In his spare time, Eric enjoys hiking, theatre, travel, and reading the Economist. 

Title: Distributed Web-Crawling using Network Coordinates
Speaker: Barnaby Malet (DoC, Imperial College)
When: Thursday 19/11/2009, 2.00, Huxley Building 218

Abstract: In this talk, we will describe our experience in the design and implementation of a large scale distributed crawling system. Specifically we will describe the implementation of a policy to assign crawlers to web servers which is termed an assignment strategy. Early research in the area focused on maintaining balanced assignments whilst reducing communication overhead between crawlers as much as possible. Later research found that by factoring proximity into the assigments, download times could be reduced significantly. These methods require additional communication with web servers to measure proximities and they are not adaptive to changes in the underlying crawler nodes (such as load fluctuations) and sudden changes in network topology.

We will describe our assignment strategy which is innovative in that it assigns Euclidean coordinates to crawlers and web servers such that the distances in the space give an accurate prediction of download times, we achieve this without any additional communication with the web servers. We will demonstrate that our method gives the system the ability to adapt and compensate for changes in the underlying network topology and in the crawler nodes, and in doing so can achieve significant decreases in download times when compared with other approaches.

Title: Content-Based Routing
Speaker: Victor Faion (DoC, Imperial College)
When: Thursday 12/11/2009, 2.00, Huxley Building 218

Abstract: Today's Internet is inherently content-based, almost all traffic is made up of users obtaining named data. The supplier of the content is not relevant, only the content itself. However the Internet was designed for a conversation communication style between two hosts and the underlying network knows nothing about the data flowing through it.

In a content-based network the flow in information is determined by the interests of consumers and the content itself. We developed a content-based router which routes messages according to their content, not a destination address.  Members of a content-based network declare their interests in the form of a predicate. Messages are injected into the network, without an intended recipient, and any interested parties receive them.

We developed a link-state content-based routing protocol based on the abstract per-source routing scheme. This scheme is based on each node calculating the shortest paths trees for all sources in the network and labelling them with the predicates of its descendants. This can lead to large forwarding tables restricting content-based routing to small networks. We implemented two algorithms that allow routers to group and simplify predicates, drastically reducing the size of forwarding tables.

Title: Transmission Scheduling in Wireless Networks with SINR Constraints
Speaker: Themistoklis Charalambous (Imperial College)
When: Thursday 5/11/2009, 4.00, Huxley Building 218, QUADS seminar

Abstract: Even though there is substantial amount of work in wireless networks scheduling, an efficient analytical solution for the transmission scheduling problem under the physical model has not been developed yet. In this work, we address the problem of finding the minimum number of time-slots (or channels) required in any given network and the corresponding transmitting powers, such that all communication requests are being processed correctly, while fulfilling specific constraints for successful transmissions that conform with the channel QoS requirements. Our methodology can be used as a benchmark for the performance evaluation of heuristic or distributed algorithms that aim for a near-optimal joint scheduling and power assignment without information about the whole network. Numerical examples and simulations are presented to illustrate the validity and performance of our proposed methodologies. 

Title: SAN-ML - A Machine-Learning Based Approach to Diagnosing SAN Misconfigurations
Speaker: Douglas Willcocks (Imperial College)
When: Thursday 29/10/2009, 2.00, Huxley Building 218

Abstract: As storage area networks are reconfigured to meet the ever-changing requirements of enterprise storage, it is not uncommon for misconfigurations to slip into production systems, often causing many hours of downtime and data loss. These misconfigurations can be prevented by validating proposed configurations against a database of so called "best practices" in an attempt to catch any errors before they are deployed to critical systems. Up until now, these best practices have been determined by hand, often by spending many hours hunting through logfiles and user submitted error reports. This method is time-consuming, expensive and imprecise. A proposed solution is to use machine learning techniques to automate this diagnostic process. Decision Trees and Inductive Logic Programming have both presented interesting results, but due to inherent limitations of either the techniques or their available implementations, neither of these approaches meet the requirements of the problem. The current approach uses HR, a theorem generation program developed by Simon Colton from the Computational Creativity Group at Imperial, and is able to produce satisfactory results. The done so far has mainly concentrated on the correctness of the generated best practices, although the performance is becoming a major bottleneck as the size of the SAN datasets grow and as such will be the main emphasis for the next stage of the project.

Title: PortLand: A Scalable Fault-Tolerant Layer 2 Data Center Network Fabric
Speaker: Andreas Pamboris (Imperial College)
When: Thursday 22/10/2009, 2.00, Huxley Building 218

Abstract: This work considers the requirements for a scalable, easily manageable, fault-tolerant, and efficient data center network fabric. Trends in multi-core processors, end-host virtualization, and commodities of scale are pointing to future single-site data centers with millions of virtual end points. Existing layer 2 and layer 3 network protocols face some combination of limitations in such a setting: lack of scalability, difficult management, inflexible communication, or limited support for virtual machine migration. To some extent, these limitations may be inherent for Etherent/IP style protocols when trying to support arbitrary topologies. We observe that data center networks are often managed as a single logical network fabric with a known baseline topology and growth model. We leverage this observation in the design and implementation of PortLand, a scalable, fault tolerant layer 2 routing and forwarding protocol for data center environments. Through our implementation and evaluation, we show that PortLand holds premises for supporting a "plug-and-play" large-scale, data center network.

Summer 2009

Title: SAMProc: A Middleware for Self-adaptive Mobile Processes
Speaker: Ruediger Kapitza (University of Erlangen-Nürnberg, Germany)
When: Tuesday 18/8/2009, 11.00, Huxley Building 217

Abstract: Software in Ubiquitous Computing is faced with a dynamic and heterogeneous environment. For tapping the environment's full potential, software must be able to adapt dynamically and react to the environment in a platform- and language-independent manner. However, developing mobile and adaptive applications is a complex and error-prone task. This talk presents the design of SAMProc, a new middleware that supports developers in creating ubiquitous applications. I will introduce the novel concept of self-adaptive mobile processes, which allows an abstract high-level specification of an application's lifecycle and its distribution aspects. Based on SAMProc, an application may dynamically migrate with an adaptation of its interface, state and implementation at runtime, while transparently and permanently remaining addressable. The SAMProc middleware uses an MDA-like approach, in which the developer creates a self-adaptive mobile process description. A middleware tool automatically maps these descriptions to our infrastructure, i.e., either self-adaptive migratable CORBA objects or Web services. The developer focuses on application logic; the tool generates migration and
adaptation code.

Bio: Ruediger Kapitza is an assistant professor leading the Distributed Systems Group at the Department of Computer Sciences, University of Erlangen. He teaches on distributed and fault-tolerant systems. Dr. Kapitza received his M.Sc. and Ph.D. degree from the Dept. of Computer Sciences, University of Erlangen in 2001 and 2007, respectively. His research interests include adaptive middleware, fault tolerance, and distributed algorithms with focus on system aspects.

Spring 2009

Title: Properties of complex networks beyond the power-law
Speaker: Shi Zhou (University College London)
When: Friday 19/6/2009, 14.30, Huxley Building 218

Abstract: It is well known that many complex networks in natura and society have a power-law degree distribution, which means a small number of nodes have very large numbers of links whereas the majority of nodes have only a few links. In this talk we discuss a number of other topological properties by contrasting two very different power-law networks, namely the Internet (at the autonomous systems level) and the scientific collaboration network (in the area of condense matter physics). The size of the Internet is enormous yet the network is very `small' in the sense that it is extremely efficient in routing data packets across the globe.  We explain that it is because the Internet has two particular structures: firstly the disassortative mixing, which means poorly-connected nodes tend to link with well-connected nodes, and vice versa; and secondly the rich-club phenomenon, where the best-connected nodes are tightly interconnected with each other forming a so-called rich-club. By contrast the scientific collaboration network exhibits the assortative mixing, which means nodes tend to link with those that are similar to themselves; and the network does not show the rich-club phenomenon. Recently we discovered that in social networks, what matters is not only how many people one knows, but also who he knows. We call it the second-order assortative mixing.

Title: Verification of Future Communication Networks
Speaker: JaeSeung Song (Imperial College)
When: Thursday 11/6/2009, 14.00-14.45, Huxley Building 218

Abstract: The introduction of future flexible networks and the wide spread of small home base stations are changing the way in which we manage network entities autonomously. These self-organizing network (SON) become larger, more complex and difficult to configure manually. Therefore new mechanisms to configure complex network systems autonomously are required. A challenging issue in this context is to enable algorithms embedded in these home base stations to avoid selecting the same physical cell identifier (PCI) as its neighbors during configuration. In this talk, I will introduce how automatic verification techniques can be adopted to check the satisfaction of self-configuration properties in a self-organising network (SON). My results show that in order to reduce the identical PCI selection problem within a SON, it is better to systematically verify protocols and algorithms designed for SONs. I will also discuss how the PCI selection problems can be analysed using model checking tools and can be avoided in the design stage.

Title: Design and Evaluation of a High-Throughput Content-Based Routing Protocol Using Unicast State and Probabilistic Encodings
Speaker: Giovanni Toffetti Carughi (University of Lugano)
When: Thursday 4/6/2009, 14.00-14.45, Huxley Building 218

Abstract: A content-based network is a content-based publish/subscribe system architected as a datagram network: a message is forwarded hop-by-hop between brokers and delivered to any and all hosts that have expressed interest in the message content. Existing publish/subscribe messaging systems, including such commonly used ones as Apache's ActiveMQ and IBM's WebSphere MQ, are capable of high message throughputs in local area settings but exhibit a degradation in performance as the broker network grows in size. We argue that this performance degradation is due the protocol they use to route messages and that the routing protocol substantially dictates the end-to-end messaging performance overshadowing the impressive capabilities of the individual brokers. We present the design of B-DRP: a content-based routing protocol that can demonstrably improve the situation. B-DRP is based on two  main techniques: a message delivery mechanism that utilizes and exploits unicast forwarding state and a probabilistic data structure to efficiently represent and evaluate receiver interests. A comparative evaluation of a B-DRP implementation with ActiveMQ is presented on a variety of topologies and for workloads representing realistic content-based applications.

Title: Resource Provisioning for Virtualized Server Applications
Speaker: Eva Kalyvianaki (University of Cambridge) 
When: Friday 29/5/2009, 13.00-13.45, Huxley Building 218

Abstract: Data center virtualization allows cost-effective server consolidation which can increase system throughput and reduce power consumption. Resource management of virtualized servers is an important and challenging task, especially when dealing with fluctuating workloads and complex multi-tier server applications. Recent results in control theory-based resource management have shown the potential benefits of adjusting allocations to match changing workloads.
In this talk I will present a new resource management scheme that integrates the Kalman filter into feedback controllers to dynamically allocate CPU resources to virtual machines hosting server applications. The novelty of our approach is the use of the Kalman filter---the optimal filtering technique for state estimation in the sum of squares sense---to track the CPU utilizations and update the allocations accordingly. Our basic controllers continuously detect and self-adapt to unforeseen workload intensity changes.
Our more advanced controller self-configures itself to any workload condition without any a priori information and improves the performance of workload-aware controllers under high intensity workload changes. In addition, our controllers are enhanced to deal with multi-tier server applications: by using the pair-wise resource coupling between application components, they provide an increased server performance improvement when facing large unexpected workload increases when compared to controllers with no such resource-coupling mechanism. We evaluate our techniques by controlling a 3-tier Rubis benchmark web site deployed on a prototype Xen-virtualized cluster.

Bio: I have recently completed my PhD in the Computer Laboratory at the University of Cambridge under the supervision of Dr. Steven Hand. My research interests include systems and most recently autonomous computing. I am generally interested in applying mathematical and statistical reasoning to address the complexity of modern systems.

Title: Towards a "functional" Internet
Speaker: Anil Madhavapeddy (University of Cambridge) 
When: Thursday 28/5/2009, 13.00-13.45, Huxley Building 218

Abstract: The Internet represents a myriad of different software programs all communicating across the planet. Unfortunately, most of these components are constructed in an ad-hoc and hence insecure fashion, leading to 'storms' of malware attacks regularly affecting millions of computers.

As a systems researcher, I investigate the intersection of modern programming languages and distributed systems in order to easily construct robust, secure and scalable networked systems. In this talk, I will describe the Melange project and the results of rewriting existing systems using modern techniques. I will then talk about the future direction Melange is taking into cloud computing, sensor networks and distributed computation.

Title: Extensible Event-Based Middleware: Achieving extensibility while maintaining security
Speaker: Ioannis Papagiannis (Imperial College)
When: Thursday 21/5/2009, 14.00-14.45, Huxley Building 218

Abstract: Complex middleware frameworks are made out of interacting components which may include bugs. Moreover, these frameworks are often extended to provide additional features by third-party code that may not be completely trusted and, as a result, compromise the security of the whole platform. Two fundamental problems that thus appear in these frameworks are data confidentiality and integrity. Some components may leak sensitive user information while others try to interfere with irrelevant data, damaging the platform's integrity.
 
In this talk, I will present my research on applying  Decentralised Information Flow Control(DIFC) on extensible middleware. DIFC is a promising label based taint-tracking technique that provides guarantees for data confidentiality and integrity. After presenting the label model that we have developed, I will outline an application of the paradigm to building a simple extensible publish/subscribe system. The system will use DIFC to hide the messages' content from an untrusted component and also to protect the privacy of the subscribers that use it. Moreover, I will present some future directions and challenges that appear on DIFC-enabled middleware frameworks.

Title: Dependable Internet-Scale Stream Processing
Speaker: Marco Fiscato (Imperial College)
When: Thursday 14/5/2009, 14.00-14.45, Huxley Building 218

Abstract: Thanks to the ever increasing availability of cheap sensor devices and network capacity, a new class of distributed applications is emerging, which we refer to as Internet-scale stream processing. It allows to process sensor data coming from sources scattered around the world over a distributed computing infrastructure. Similar to the ease of relational queries in DBMS, stream processing systems allow users to access and manipulate distributed data streams through declarative queries. However, the scale of an Internet-wide system poses substantial challenges when it comes to providing a dependable service. Any such system must racefully handle the failure of network links and processing hosts while managing a large pool of CPU and network resources. The topic of this talk is my work on DISSP (Dependable Internet-scale Stream Processing), an infrastructure for distributed stream processing focusing on optimizing the dependability of the system.
Even in case of insufficient resources or failure, it aims at providing approximate meaningful results. I present my research on developing a new quality-centric data model which allows to estimate the impact of the data loss in the processing, giving the system a way to gracefully degrade the quality of the results while constantly providing feedback on the achieved quality of service. I'll also talk about the implementation of this model in Borealis, a previous stream processing engine, and about the its applications for replication, operator placement and load-shedding. 

 Title: A generic framework for P2P sampling and monitoring
Speaker: Cyrus Hall (University of Lugano)
When: Wednesday 29/4/2009, 16.00-16.45, Huxley Building, LT311 (Note change of date and location due to exams)

Abstract: The development of peer-to-peer (P2P) networks has enabled the decentralization of applications and services across heterogeneous commodity hardware.  However, monitoring and understanding the dynamics of deployed peer-to-peer systems remains an very difficult problem.  The size and dynamics of widely distributed systems make it impossible or undesirable to maintain accurate and complete membership information in a centralized fashion, while decentralized gossip-based peer membership algorithms often exhibit biased peer selection.  One technique being explored to overcome these difficulties is peer sampling.  Peer sampling is the process of selecting a peer at random from a given distribution (typically uniform) over all members of a network.

In this talk I will present a modular and generic peer sampling algorithm called Doubly Stochastic Converge, based on random walks and the properties of doubly stochastic matrices.  DSC forms one part of a peer sampling framework we are building to enable more accurate monitoring and maintenance of deployed P2P networks.  I will present details of this framework, as well as a generic and transparent message-aggregation scheme that can reduce the overhead associated with sampling.

Title: Clustering Weblog Posts Using Tags
Speaker: John Ayres (DoC, Imperial College)
When: Thursday 2/4/2009, 14.00-14.45, Huxley Building, Room 218

Abstract: Content-based routing is a generalised multicast messaging service hereby multicast groups are defined by the interests of their members rather than by explicit designation.  An often cited use of such networks is personalised news distribution whereby subscribers receive only the news of interest to them. This area has received little practical attention. To apply the routing scheme to this domain requires a distributed decentralised news source and a set of readers with varied interests. For these reasons mainstream news sources are unsuitable and we turn to the rapidly growing network of blogs. The worldwide network of blogs produces around 900,000 posts per day, a suitably large workload for the scheme.

In this talk we outline how the weak (disconnected and sparse) link structure, short content and ephemeral nature of blog posts exclude the use of traditional information retrieval techniques for document classification. We present an investigation into the effectiveness of tagging, an increasingly popular component of 'web 2.0' applications, as a feature for clustering posts around a common theme. By using these themes to generate subscriptions we show how readers' interests may be expressed in a content-based network for efficient news distribution.

Title: Content Based Routing Structures: From Trees to Mobile Approaches
Speaker: Matteo Migliavacca (DoC, Imperial College)
When: Thursday 26/3/2009, 14.00-14.45, Huxley Building, Room 218

Abstract: Content Based Publish Subscribe is a communication and interaction paradigm that is quite promising to build complex software architectures. In particular in its distributed incarnation, it can be used to decouple application components in dynamic environments, e.g. in peer to peer applications or in wireless mobile settings. However dynamism in these systems pose challenges in designing efficient routing structures. In this talk I will survey some popular approaches used in these systems, from tress, to graphs, to structures more adapt to mobile settings. I will also cover some enhancements that I developed while working at DEEPSE group at Politecnico di Milano.

Bio: Matteo Migliavacca is a Research Associate at DSE group working on the SmartFlow project after obtaining a Ph.D. degree at Politecnico di Milano with a thesis on Middleware Services for Large Scale Dynamic Distributed Systems. His main research interests are in Distributed Systems in the areas of routing, adaptivity, context-awareness and programming abstractions.

Title: Distributed Path Clustering for Internet-Scale Detour Routing
Speaker: Thom Haddow (DoC, Imperial College)
When: Thursday 19/3/2009, 14.00-14.45, Huxley Building, Room 218

Abstract: For a large proportion of default end-to-end routes over the Internet there exists an alternative route that could improve on its quality. Since the Internet, in its current incarnation, does not expose an interface for explicitly specifying routes we can only access these alternative routes by redirecting traffic via tertiary endpoints -- "detour nodes". In order to exploit this "slack" in Internet routing these detour nodes must first be discovered through Internet measurements, but existing systems suffer from scalability concerns resulting from the need to take a large number (O(n^2)) of these measurements, or delegate this measurement to secondary services.

In a previous talk, Sing Wang Ho described "Path Clustering", an approach which allows detour nodes to be identified using a much reduced number of measurements. Here I will discuss this in the context of routing overlay networks and introduce "Distributed Path Clustering", a gossip-based distributed algorithm enabling the path clustering method to be used within large scale distributed systems.

Title: Query Optimization in Distributed Data Stream Systems
Speaker: Quang Hieu Vu (DoC, Imperial College)
When: Thursday 12/3/2009, 14.00-14.45, Huxley Building, Room 218

Abstract: Distributed data stream systems (DDSS) have become popular of late. Typical examples of DDSS include a network monitoring system that analyzes network traffic to detect attacks of intrusions, a banking management system that studies financial transactions for fraud detection or a health care management system that monitors the behavior of patient/elderly people to signal emergency attention when necessary. In DDSS, query optimization is a challenge due to the large scale of the system, the lack of global knowledge, and the dynamic nature of the system.

In this talk, I will present a combination of two solutions for query optimization in DDSS. The first solution is to employ indices on query operation. The purpose of this solution is to allow the system to reuse computation of common sub-queries inside queries. The second solution is to adaptively optimize query execution plans based on an economy approach. In this solution, computation nodes offer prices of data streams they can provide and query operations they can perform based on their available resources and the query execution plans are optimized to select the cheapest supplier from a set of suppliers that can provide the same data stream/query operation.

Title: Building Large-scale Distributed Systems with Network Coordinates
Speaker: Peter Pietzuch (DoC, Imperial College)
When: Thursday 5/3/2009, 14.00-14.45, Huxley Building, Room 218

Abstract: Network coordinate systems embed round-trip time measurements between Internet nodes in a virtual metric space. Initially, network coordinates were proposed as way of estimating latencies between nodes in peer-to-peer systems. However, their applicability is more general and they can act as a building block for locality-aware Internet-scale distributed applications.

In this talk, I will give an overview of our work on designing real-world network coordinate systems and using them to solve problems in large-scale distributed applications. First, I will describe techniques for making network coordinates more practical by increasing their stability and accuracy. I will then talk about different uses of network coordinates that we explored: to provide a general-purpose routing overlay, to improve the locality-awareness of BitTorrent and to map clients to servers in a content delivery network.

Title: Ukairo: Improving End-user Experience using Detour Routing
Speaker: Sing Wang Ho (EE, Imperial College)
When: Thursday 26/2/2009, 14.00-14.45, Huxley Building, Room 218

Abstract: The Internet has the capability to deliver vast amounts of information between end hosts.  However, the Internet delivers packets without considering the specific needs of applications.  For example, video and voice applications demand low latency and low jitter connections whereas file downloads require high bandwidth. In the Ukairo project, we construct an overlay network where packets are detoured to other nodes within the overlay to find better paths with respect to application-specific metrics. This talk will be split into two parts: First I will describe how to measure metrics of Internet paths, then I will discuss how detours are found using these measurements.

Title: Network Verification in Future Networks
Speaker: JaeSeung Song (DoC, Imperial College)
When: Thursday 19/2/2009, 14.00-14.45, Huxley Building, Room 218

Abstract: The Virtual Centre of Excellence in Mobile and Personal Communications (MobileVCE) is a collaborative partnership engaged in industrially-led, long term networks research. In the past, MobileVCE carried out research on Beyond 3G, Ubiquitous Services and Personal Distributed Environments. Recently MobileVCE has decided on 'Flexible Networks' as the theme for their latest Core 5 research programme. This programme will investigate areas, such as network virtualisation, self-* properties and network coding, that are important for flexible future networks

Any such flexible network must be trustworthy and guarantee correctness of its behaviour. Therefore novel network verification techniques are an important requirement of future networks. In this talk, I will introduce the aims and features of future flexible networks. I will also describe the role and methodologies of network verification for self-organizing systems. The 3rd Generation Partnership Project (3GPP) has standardized a specification for a self-organizing network consisting of Home Node B (HNB) cells. Many HNBs will be deployed in future mobile networks, which makes the verification of self-configuration of HNBs an important example for research in this domain.

Title: Dependable Internet-scale Stream Processing
Speaker: Marco Fiscato (DoC, Imperial College)
When: Thursday 12/2/2009, 14.00-14.45, Huxley Building, Room 218

Abstract: The wide availability of sensing devices scattered around the world and the increasing capabilities in computing power and network capacity are giving rise to a new class of distributed applications, which we refer to as Internet-scale Stream Processing. Similar to the ease of relational queries in DBMS, stream processing systems allow users to access and manipulate distributed data streams through declarative queries. However, the scale of an Internet-wide system poses substantial challenges when it comes to providing a dependable service. Any such system must gracefully handle the failure of network links and processing hosts while managing a large pool of CPU and network resources.

The topic of this talk is DISSP (Dependable Internet-scale Stream Processing), an infrastructure for distributed stream processing focusing in optimizing the dependability of the system. This means that in presence of failure or in the case of insufficient resources the system will provide approximate results, while constantly providing feedback on the achieved quality of service.

Title: SmartFlow, Event Based Middleware for Healthcare Applications
Speaker: Ioannis Papagiannis (DoC, Imperial College)
When: Thursday 5/2/2009, 14.00-14.45

Abstract: In this talk I will present SmartFlow, an extendable event based middleware framework for healthcare applications. SmartFlow is a 3-year ESPRC funded project that started this October and is carried out by Imperial, Cambridge and CBCU. It aims to interconnect various NHS systems and answer the need for electronic solutions that most healthcare providers demand in order to improve their services to the public. SmartFlow will be highly reconfigurable: it will allow multiple applications to build upon the core functionality it offers in order to support the interconnection of various systems used across the NHS. Since confidentiality and data integrity are core demands for every application in this domain, the main focus of the work already done explores how the paradigm of Distributed Information Flow Control (DIFC) can be applied in our system. In this talk I will showcase how DIFC can avoid encryption while on the same time allows the system administrators to reason about the confidentiality and integrity of the data that applications built on top of SmartFlow process.