for telephony, conferencing … • Often used over best-effort UDP/IP networks • RTP provides playout timing and packet sequencing • Reception quality feedback every few seconds (RTCP) • RTCP provides higher-level summary feedback ( jitter, fractional loss, RTT, last packet received packets sent, octets sent)
do not require applications to do anything • Keep the RTP/RTCP flow properties – i.e., do not confuse middleboxes, monitors, … that see packets on a single path only
Sort paths based on • Characterize paths based on loss and discards* – No Losses or discards: non-congested – RTCP reports losses/discards: mildly congested – 3 consecutive RTCP reports of losses/discards: congested • Ignore paths where RTT > max_end_to_end_delay [CB] http://tools.ietf.org/html/draft-ietf-avtcore-rtp-circuit-breakers *discards: packets that arrive late are discarded throughput delay
– Available throughput – High – lowest loss rate. • All packets belonging to a video frame are sent on the same subflow • Fractional Distribution for each path is Path_ RR Path_ RR ∑ × Media_ Rate throughput delay
in sec 0 0.2 0.4 0.6 0.8 1 0 50 100 150 200 250 max_end_to_end_delay = 400ms optimum_end_to_end_delay = 150ms 50 ms 150 ms 250 ms Use path for ARQ, I-frames Increasing traffic distribution