Advanced messaging technologies

Message-oriented middleware (MOM) has been used since the late 1980s as a flexible way to build distributed, heterogeneous systems. Asynchronous message delivery, data-centric addressing/routing and transformation capabilities are the characteristic features of MOMs that provide time and space decoupling and allow the integration of heterogeneous systems.

The Advanced Messaging Technologies (AMT) research group works on technologies to extend the reach and applicability of MOM-based integration into new areas outside the traditional IT enterprise field.

Real-time and QoS messaging

Many real-world-aware solutions require the timely delivery of information between the distributed edge components (i.e. sensors and actuators) and more centralized business systems. In this project, we are developing a publish/subscribe messaging system, where timely delivery of messages with low latency and high data rates is addressed by the core architecture. This architecture has the ability to allocate resources to topics through the system's messaging scheduler, which allows message delivery to be prioritized dynamically based on topics. To meet the requirement for high-volume throughput with low latency, the architecture exploits the parallelism provided by today's multi-core platforms.
More information.

Algorithms and data structures for parallel systems

Multi-core processors have become the standard architecture for general-purpose machines. Traditional lock-based data structures and algorithms do not scale well with the number of CPU cores. We are investigating lock-free approaches that scale with the number of processors for use in high-performance messaging systems.

Verifying the correctness of parallel algorithms is a very important but nontrivial task. We are developing a method to prove program traces of parallel code on linearizability. The results of the verification process allows efficient debugging and, as a result, yields higher-quality code.
More information.

Sensor network and sensor network integration

In this project, we are considering an end-to-end solution where sensors and actuators can connect via a message-oriented middleware to enterprise backend systems such as application servers. We are developing publish/subscribe protocols that operate on a multitude of transport protocol stacks — even very limited ones — such that they can be implemented on very low-end devices. Communication efficiency can be increased further by taking the needs of applications into account. This is done by pushing intelligence for evaluating and analyzing sensor data into the sensor network such that only relevant information is exchanged, thus allowing both bandwidth and energy consumption to be reduced. We are developing a communication framework that evaluates the tradeoff of distributed processing and communication as part of a sensor data evaluation algorithm on smart sensor nodes. The framework transparently implements the optimal distributed processing strategy for a given goal, e.g. optimal battery lifetime.
More information.