The ETX Path Metric for multi-hop wireless

Douglas S.J. Dr Couto, Daniel Aguayo, John Bicket, Robert Morris, "A High-Throughput Path Metric for Multi-Hop Wireless Routing"

Computation of ETX metric

• total metric=sum of the links
• for one link, metric 1/(forward delivery ratio*reverse delivery ratio) (probability that packet was received, probability ACK was received)
• estimation through 1 probe/sec broadcast by each node with jitter to avoid collision
• gives expected value of total transmissions (with retransmissions) of packet along that path
  

Objectives (motivated by measurements of throughput through different routes)

• accounts wide range of link loss ratios
• accounts asymmetric loss ratios
• accounts interference between successive hops of multi-hop paths (one can't send til one has completly received)
• tends to minimize spectrum use, overall system capacity, energy

Results

• Tested with two modified routing algorithms: DSDV and DSR in office environment
• significant higher throughput through multi-hop routes
• algorithm does not account different loss ratios for different packet sizes

I was wondering whether (when there several packets on its way) the network actually becomes a pipeline which can work (at least) partially parallel and therefore the ETX metric does not optimal for max throughput.
I also still do't understand why the graphs in figure 6 cross: this clearly is a contradiction to the fact that the best route should perform equally or better than any other route, so why does its graph suddenly lie above another route?