Search Search Search Evangelos Kotsovinos home page

Evangelos Kotsovinos
Research
   

My main research interests are in the broad area of distributed systems. Before joining Morgan Stanley, I used to lead the XenoServers project. My publications and patents are organised by subject below.
Virtualization and Global Public Computing: A next generation distributed computing paradigm that, in one phrase, allows ``anyone to run any code anywhere''; servers scattered across the globe make computing resources available to all members of the public, not just cooperative scientists, and to all applications, not only well-behaved scientific experiments. They do so in exchange for money; users are ultimately charged for the resources their applications use on the servers. The XenoServer Open Platform substantiates that vision.

Publications:
Federated Resource Management: Controlling resource consumption poses a major challenge for Grid computing, peer-to-peer systems and utility computing platforms. Problems stem from hardware heterogeneity and diversity in the resource management policies desired by the different organisations involved. For instance, the owner of a machine, the institution hosting the machine, the user requesting resources on a machine and the user's own administration often have very different goals and consequently different policies on the portion of resources that should be allocated in any given setting. I have developed Role-Based Resource Management for allowing federated policies to be expressed and flexibly combined.
 

Publications:

Trust and reputation management: One of the major ongoing problems, as well as common reasons of failure, in public distributed computing is the difficulty of establishing trust relationships between the parties involved, due to the heterogeneity of participants, their disparate goals, and independent nature. I have been working on XenoTrust, an event-based reputation system comprising a publish/subscribe methodology for the scalable storage, retrieval and aggregation of reputation information. I have devised Pinocchio, a system that probabilistically assesses the honesty of participants, and provides explicit incentives for participation.
 
Publications:

Systems support for ubiquitous and pervasive computing: The emergence of ubiquitous and pervasive computing as a new promising area of computer applications has raised a number of systems research challenges, due to the complexity, transience, and scale of such environments. I am looking at novel systems technologies to support ubiquitous computing, such as replic8, a smart service replication platform for ensuring high availability at a low cost, and su-chef, an architecture for the automatic coordination of smart home environments.
 

Publications:
Distributed and operating systems: Systems that run on large numbers of machines around the world, either to split the computational load of demanding applications or to allow sharing files and other resources, have become increasingly popular in the last few years. I am mainly interested in systems support and design issues for large-scale applications, such as peer-to-peer applications and distributed file systems, and the associated effects on scalability, performance, fault-tolerance, and flexibility. I am also interested in looking at the properties of different operating system architectures, and how design decisions affect the penetration of the markets they are targetting - in terms of user groups, applications, and long-term sustainability.