Slashdot Mirror


Reviving Ricochet: Better Than WiFi?

renard writes "Slate is carrying a column by Brendan Koerner arguing that reviving the Ricochet city-wide wireless network infrastructure would be a better idea than blanketing the nation/world with 802.11-ish WiFi. He reviews all the usual silly reasons why Metricom, the original owners, were unable to make a go of it, and makes a good case that things may go better the second time around."

1 of 161 comments (clear)

  1. Some significant research was done in this area by ekrout · · Score: 1, Redundant

    A group of college students reviewed Richochet's wireless network, with the following conclusions being made:

    We presented the results of several experiments performed to evaluate the performance of Metricom's Ricochet network. We investigated the performance of the experimental deployment as part of the Daedalus project, as well as the commercial service. Our main conclusions and observations are summarized below.

    • TCP throughput is between 15 and 30 kb/s, across a wide range of packet and socket buffer sizes. In the commercial service, peak throughputs are observed for a packet size of 512 bytes and socket buffer sizes of 4-8 KBytes.

    • TCP performance suffers from long idle periods caused by reverse channel contention; large variations in round-trip time estimates (between 250 and 5000 ms) greatly increase loss recovery times and significantly degrade performance.

    • Maximum UDP throughput is between 50-58 kb/s, when there is no reverse channel contention; parallel UDP streams in the same direction show a roughly linear performance degradation.

    • UDP jitter under lightly loaded optimal conditions are between 10 and 30 ms across the range of supportable transmission rates. This implies the viability of real-time video and audio transmissions at low bandwidths.

    • Measurements of interactions with the WaveLAN device show that the performance of the Ricochet network does not degrade as much as the WaveLAN network when simultaneous transmissions occur.

    • The performance of the experimental deployment is roughly similar to the commercial service, implying that improvements made to the former are likely to improve the performance of the latter as well.

      The main page is at this site.

    --

    If you celebrate Xmas, befriend me (538