2017/2018 Edition

Gossiping in Distributed Systems / Web-based Collaborative Decentralized Systems

Lecturer: Prof. Maarten van Steen (Vrije Universiteit, Amsterdam, The Netherlands).

About the lecturer | Course Summary | Scribe notes

About the lecturer: Maarten van Steen received a Masters degree in Applied Mathematics from Twente University, and a PhD in Computer Science from Leiden University. He is currently a full professor at VU University in Amsterdam where he teaches systems-oriented courses and conducts research on large-scale distributed systems, with an emphasis on systems where nodes can take decisions on only locally available information. These include wireless systems (notably large-scale sensonets) and wireline systems (such as decentralized grid infrastructures and traditional peer-to-peer systems). He as an author of well-known textbooks: "Distributed Systems: Principles and Paradigms" (with A. Tanenbaum) and "Computer and Network Organization: An Introduction" (with H. Sips). He advised 11 PhD students.

Course summary:
Distributed systems continue to face difficult scalability problems while expanding in number of components as well as the communication latencies between components. This has led to an increased interest in fully decentralized solutions in which nodes take decisions based on only locally available information. In particular, gossiping by which (meta-)information is disseminated or aggregated across the network forms an important technique to come to such decentralized solutions.
In this mini-course, we concentrate on (the role of) gossiping in distributed systems. Gossip-based solutions show a great deal of emergent behavior, and it is often unclear what the relationship is between a specific configuration of a gossiping protocol and its behavior. Likewise, we often see a difference between what theory would predict about a protocol, and what happens in practice, notably when concentrating on extra-functional behavior such as reliability, robustness, and reactions to changes in node membership.
The course will be roughly divided into two parts: foundations, in which theory and experimental findings are discussed; and applications, covering aggregation, structure management, and wireless systems.

We are currently witnessing two, seemingly diverging trends in distributed systems: one, in which fully decentralized peer-to-peer computing is deployed for exchanging information between end users, and another one in which we see centralization of Web services take place through very large data centers. In this mini-course, we consider where the two approaches meet, namely in Web-based (collaborative) decentralized systems, and content distribution networks in particular.
To meet scalability requirements, replication of Web content plays a crucial role in these systems. We first concentrate on the issues that need to be considered when applying replication techniques for the Web, followed by a presentation of various experiments that show that we need a myriad of (application dependent) solutions in order to actually build scalable systems. In doing so, it will uncover more fundamental problems for which no general solutions are yet known, but which need to be tackled in order to make any significant progress.


Some additional material here