Improving transport design for WARP SDR deployments

2014 
Software-Defined radios (SDRs) are a popular platform for developing and implementing wireless protocols. Their basic architecture consists of radio front-ends hosted on an FGPA board, and a back-end processing host for running bulk of the signal processing in software. The two components are bridged, usually by an Ethernet or PCIe interface that transports the radio samples. In addition to the processing delay in software, SDRs may experience a non-negligible transport latency, for example, due to the limited Ethernet bandwidth. Wireless-Access Research Platform (WARP) is one such SDR platform that has recently gained a lot of attention. Research prototypes deploying tens of WARP radios over the Ethernet have become a familiar sight. WARP's transport design, however, is inefficient due to its linear increase in transport latency with the number of radios. We propose modifications to improve the current design. First, we utilize functional parallelism to run the read/write operations of multiple WARP radios concurrently. Second, we propose a high-bandwidth link at the host in order to support the combined transfer rates resulting from the parallel transport to/from the radios. As a result, we achieve a significant reduction in the transport latency by scaling back the linear increase to a constant overhead.
    • Correction
    • Source
    • Cite
    • Save
    • Machine Reading By IdeaReader
    12
    References
    3
    Citations
    NaN
    KQI
    []