Diploma Thesis DIP-3594

BibliographyBosch, Patrick: Solving the Content Routing Problem with Coupled Constraint Oriented Programs.
University of Stuttgart, Faculty of Computer Science, Electrical Engineering, and Information Technology, Diploma Thesis No. 3594 (2014).
87 pages, english.
CR-SchemaC.2.1 (Network Architecture and Design)
C.2.4 (Distributed Systems)
G.1.6 (Numerical Analysis Optimization)
Abstract

Publish/Subscribe and especially content-based publish/subscribe are widely used for communication between distributed components. They offer efficient bandwidth usage through forwarding events only to subscribers which are interested in the content of the event. Forwarding is done by filters which identify the content of the event.

Previous implementations of such systems relied on a distributed set of brokers to match and forward the events accordingly. However, the advent of software-defined networking gave us the possibility to implement the system directly on the network layer. The overhead of the application layer can be prevented and the filtering process can be implemented more efficiently. Previously, the costs for filtering were a major disadvantage for content-based publish/subscribe systems. Such a new filtering approach with implementation on the network layer was presented in initial work and it was shown that line-rate performance could be achieved. For this approach the power of software-defined networking could be fully harnessed by implementing forwarding flows that handle the filtering as well as the forwarding. Although, this approach offers a possibility for a brokerless publish/subscribe system, the task to compute and optimise routes that connect publishers and subscribers is not solved adequately.

To understand the problem of route optimisation in such a system and the problems that are incorporated in it, this work divides the overall problem into multiple smaller problems and offers solutions for these smaller problems. A framework is presented and discussed with regard to achieving a complete solution which incorporates optimisation problems as well as transformations and mechanisms that ensure that each part is independent from the other parts. The performance as well as the results the framework produces are discussed after its presentation.

Full text and
other links
PDF (1803470 Bytes)
Department(s)University of Stuttgart, Institute of Parallel and Distributed Systems, Distributed Systems
Superviser(s)Koldehofe; Boris
Project(s)aks
Entry dateJuly 17, 2014
   Publ. Computer Science