The Real Time Messaging activity focus is on exploration of challenges and solutions in the field of predictable data delivery in large scale systems and systems with unpredictable elements. The software platform for implementation of the research ideas is called TransFab.
TransFab stands for Transport Fabric - a high-performance and predictable communication layer for real time middleware. It runs on top of networking infrastructure and implements a set of services that allow to create a messaging solution with explicitly enforced limit on time of end-to-end data delivery. TransFab carefully manages its internal resources during the runtime, and exchanges resources with external entities, in order to support the end-to-end latency goals in large-scale distributed systems.
A key target is compliance with requirements in military and other industries that need highly predictable transport solutions, with rich set of qualities of service.
Another focus is high throughput and low latency properties that satisfy the most demanding requirements in the market, specifically the financial front-office industry, and make it competitive with the best performing solutions in the field.
TransFab is a messaging transport technology that accepts messages from an upper layer and processes them for fast and reliable delivery to a receiver or set of receivers, either on unicast or multicast transport. TransFab interfaces and utilizes real time support in underlying platforms, such as RT OS or real time Java, and differentiated or integrated services support in networking platforms. The internal resources managed by TransFab are memory buffers required for message processing/queuing/reliability, and bandwidth. Other tools at TF disposal are the thread prioritization by OS, CPU shares and int/diff/rt QoS by underlying network. TF uses these resources dynamically, as a set of instruments to achieve the goal of fast and timely data delivery.
The ongoing research of the Real Time Messaging team builds upon the existing TransFab platform, and extends it to address the following exploratory challenges:
- Predictable task execution and communication resource management in very large scale distributed systems
- Real time data routing in mobile ad-hoc wireless networks
- Methods for high availability in real time middleware