The linear experiment

• OBJETIVES: study predictable behavior and validate dual mode.

• A linear array of terrains.• Only the leftmost station sends.• Many messages are scheduled in the

beginning (attention: this affects latency).• Only the rightmost station receives.• We also want to see the differences

between single mode and dual mode.

The linear experimentTopology

Line_2

Line_3

Line_4

AP

Client

Linear experiments results

ThroughputDelay (avg)Jitter

Line_232.7%3098 ms1754 ms

Line_2 dual38.9%2617 ms1468 ms

Line_338.7%3574 ms2003 ms

Line_3 dual62.4%2251 ms1222 ms

Line_438.6%3604 ms2003 ms

Line_4 dual61.3%1669 ms869 ms

The linear experimentResult analysis

• We find the smallest throughput in line_2, where all the transmissions affect each other, a message needs three hope to reach the destination, thus the throughput has to be very close to 33% (we have got 32.7%).Dual mode does not improve, but there is still collision between the client transmissions.

The linear experimentResult analysis 2

• In Line_3 and Line_4 there is no interference between clients and the interference between APs is only to the direct neighbors.

• The difference between dual mode and single mode is because in this cases the message that is sent can reach the end of the pipeline without interference if messages are sent every cycle!!!

Station 0

Station 1

Station 2

Station 3

Station 4

Station 5

Station 6

Station 7

For a messages of length 60, the graph on the left shows that in double mode the messages can advance in parallel in the pipeline to the following station requiring not much more time than 60.

(the graph does not consider RTS,CTS and backoff after succeeded transfer).

The single terrain experiment

Continuously modifying parameters.

The single terrain experiment

• This experiment lets us evaluate what is the local behavior of the protocol by running with only one terrain.

• We will run with Poisson distribution.• Objective: Discover how the throughput

varies with the number of stations (who have the same message distribution)

• Objective: Is the total throughput dependant to stress??

Results for non stop

These two got much x10 messages and ran more time

Results for stop on time

The single terrain experimentConclusions

• First of all we see that when the simulation runs longer time as all the stations throughput converges to the average.

• The throughput of all the network is not too much under 50%, which would be the throughput for perfect communication (a message needs two hops to arrive at destination).

The single terrain experiment Conclusions(2)

• Beyond the point that all the resources of the network start to be used the stress at the network does not seem to affect the total throughput significantly.(There is a big impact in the latency which and jitter which we have not analyzed in the previous graphs).

• Possible explanation: For the current ratio between the length of a message and the length of the RTS we see not too much collisions.

Wavion vs competition

Mesuring the throughput of wavion long range Aps.

Wavion vs competitors

• Wavion reaches a longer range (we will consider the range to be the double), that means that the terrains will have an area as four times as bigger than the terrains of the competitor and we should expect the area to include four times the amount of clients.

others

Advantages and disadvantages

Wavion networks contain less Aps, this means less internetwork traffic and less collisions.

However, in our experiments Wavion will contain more clients per AP thus creating more local collisions.

Also, Wavion does not support Dual mode that some of its competitors support.

Description of the experiment

• Different scenarios with a reasonable geometry will be presented.

• Four stations per Wavion AP have to be positions.

• In the competitors scenario we will divide each terrain to four.

• We will run the competitors scenario both with and without dual mode.

One Wavion terrain scenario(with strong Ethernet)

• Most of the messages came from the Ethernet, first we calibrated the messages to get a maximum throughput on Wavion.

Ethernet Arrival time Ethernet throughput

50014%

40017%

20025%

15026%

7034%

3035%

1537%

Comparing the results for the 70 ms arrival time

Comparing the results for the 70 ms arrival time

The Event Driven Mechanisms

• The events are entered into the EventCollector Data Base with 2 keys:– Primary Key : Time Event– Secondary Key : Priority

This way we assure pullout in the order of this keys. First by the time of event and secondary this enables us a flexible inner order of the events with the same Time Event

The Event DrivenTopology

EventCollector) TreeMap<Double,MultiTreeMap> )

Key = 40.6 Key = 80.6 Key = 85.3 Key = 90.7

timeKeyEvent) MultiTreeMap<Integer,MultiTreeMap> )

Key = 1 Key = 2 Key = 3

)Set <Events> ) )Set <Events> )