Buffer schemes for runtime reconfiguration of function variants in communication systems

This contribution introduces distributed buffering schemes for runtime reconfiguration in adaptive processing systems, e.g., real-time streaming media applications. With dynamic reconfiguration, the area-cost of field-programmable logic can be reduced by reuse of area resources, and potential for adaptive signal processing techniques can be enabled. The challenge with runtime reconfiguration is the reconfiguration latency. Given the limitations regarding reconfiguration latency with traditional approaches, we propose distributed buffering schemes, to hide latency for fast adaptive systems. To verify our approach, we do analytic and simulative modeling. The results show that our approach can guarantee zero loss-rates and reduce the additional delay (reduction of delay over 65% with estimated application profiles) in cases, where the traditional approach would not reach a desired Quality of Service. Dependent on the worst case application profile, it enables potential for runtime reconfiguration of adaptive signal processing under real-time constraints. Finally, we derive an architecture template with implementation details.