Dynamic Reconfigurable Packet Distribution Unit for Embedded Systems
Elena A. Suvorova, Valentin Rozanov · 2019 Wave Electronics and its Application in Information and Telecommunication Systems (WECONF) · 2019
The requirement of mode change (change of task's set) and faults mitigation is very actual for the most modern embedded systems. Typically, an embedded system is exploited during several years. New tasks (applications), new versions of transport protocols or new transport protocols could be developed during this time. Possibility of these new features support without change of equipment is very important today. Packet flow distribution unit is used in many embedded Systems-on-chip (SoC) and Network-on-chip (NoC) for distribution of packet flows between units of network layer and units of transport layer (for example, different transport protocol controllers or Direct Memory Access (DMA) channels). Support of new rules for packet distribution is required, when new applications and/or new transport protocols are added to system. Therefore, the packet distribution unit should be dynamically reconfigurable to meet these requirements. Today the development of dynamically reconfigurable components for embedded systems typically based on FPGA technology. The density of blocks on FPGA is essentially less, than in ASIC with the same design rules. As result FPGA implementations typically have essentially larger power consumption and essentially worse timing parameters, than ASIC implementations, the area of component on FPGA is essentially large than the area of same component on ASIC. These factors essentially constrain the application area of FPGA based dynamically reconfigurable systems. Therefore we consider the existing methods of dynamic reconfigurable systems development for ASIC, evaluate its suitability for packet distribution unit realization in this paper. We propose an approach to development of dynamically reconfigurable packet distribution unit that corresponds to requirements specific for embedded systems. In the paper we present several examples of the proposed approach using. We evaluate reachable parameters and overheads for these examples. These results are suitable for developed systems.