Media transfer over an IP network suffers end-to-end latency specifically from buffering, especially at network nodes. Current quality of service (QoS) Solutions include traffic engineering and over-provisioning, but these complicate the system, increase power requirements and add cost. This research aims at a cleaner approach to packet based real-time media.
In collaboration with Ninetiles, Cambridge, UK.
The networking of live audio for professional applications typically uses layer 2 based solutions such as AES50 and MADI utilising fixed time slots similar to Time Division Multiplexing (TDM). However, these solutions are not effective for best effort traffic where data traffic utilises available bandwidth and is consequently subject to variations in QoS. Audio can also be sent over Internet Protocol (IP), but the size of the packet headers and the difficulty of keeping latency within acceptable limits make it unsuitable for many applications.
Flexilink is a technology for a unified low latency network architecture that supports both time deterministic and best effort traffic towards full bandwidth utilisation with high performance routing/switching.
The research was initiated by John Grant of Ninetiles, Cambridge, UK. Ninetiles also developed the Aubergine technology to deploy the Flexilink transport.
DMTLab are continuing development and evaluating the technologies against other industry solutions.
There are a number of applications where Flexilink can provide improved performance including studio-based media production, networked media distribution, medical information distribution, etc.
Other research in the lab has addressed the feasibility of applying the technology to compressed video media for production and distribution.
Evaluation has addressed the low latency requirements and assessed latency (i.e. per network node) for a range of mixed traffic types - synchronous media traffic and asynchronous best-effort traffic.
Results have demonstrated a technology which handles a mix of traffic types (real-time media, best effort) with lower latency than established industry competitors.
Work is ongoing to introduce the technology into initial deployments. This is a major issue since the technology requires an end-to-end path to achieve required user latencies.