Discovery and signaling enhancements for data-centric publish-subscribe environments

Abstract

Current uses of the Internet are mostly built on request-response exchanges between pairs of computers. This model has been very successful at providing shared remote access to resources and point-to-point communications for applications including email, remote file and database access, web access, etc. The advent of ubiquitous Internet connectivity and cheap network-connected devices has greatly expanded the communication needs of applications using the Internet. For these new kinds of applications the request-response communications model, which is highly coupled to the location of the information, is cumbersome and limited whereas other communication models, such as publish-subscribe, provide a more natural fit. In contrast to the request-response interaction model, in which a client sends a request and a server returns a response, in the publish-subscribe model information producers (publishers) are decoupled from information consumers (subscribers) via the publish-subscribe service. One of these services is Data Distribution Service (DDS), a standardized solution for data-centric publish-subscribe data distribution. This Thesis describes the design and evaluation of a set of discovery and signaling enhancements for Data-Centric Publish-Subscribe (DCPS) systems. In particular, this Thesis focus on solving the following open issues of DDS: data-spaces interconnection, data transformation, Quality of Service (QoS) adaptation, universal DCPS entity and data-content naming, end-to-end session establishment, and discovery scalability in large deployments. We have divided our solution in three parts. First, we propose the DDS data-space Interconnection Service (DDS-IS), a fully DDS-compliant service for DDS data-spaces interconnection which also provides transparent data transformation and QoS adaptation. We demonstrate that the DDS-IS can improve the scalability of DDS systems by reducing the network traffic load between the interconnected data-spaces. Secondly, we define a mechanism for universal identification of DCPS entities and data-content, and we propose the rationale and design of a protocol for session signaling in DCPS environments. This protocol allows applications to perform DCPS entity and data-content discovery in medium-scale environments. It also allows the creation of sessions between nodes, which eases the creation of DCPS routes and enables the enforcement of QoS at end-to-end level. Finally, we propose a scalable Peer-to-Peer (P2P) solution for performing content discovery and data transfer. This solution is based on the REsource LOcation And Discovery (RELOAD) framework, one of the latest standards of the Internet Engineering Task Force (IETF). This solution provides rendezvous function for locating sensors, actuators, and applications that share a common data-space or topic of interest using a DCPS approach. Once the rendezvous is complete, nodes use existing RELOAD features for exchanging information. We demonstrate the proposed solution scales well for large deployments and it is robust against node failures.