Pre-computed sets of checksum packets of constant degree with various size factors

 

EG

2006-04-10

 

For a set of K source packets with a method described in 060406-setof-checksums [ ch | us ] and in 060407-setof-checksums [ ch | us ] one can generate  checksum packets of constant degree, where . In our examples , degree is equal to 7 and the factor m is from 2 to 6. For each size factor m we correspondingly generated 500 distinct sets of checksum packets (2’500 pre-computed sets). The corresponding 500 sets are sorted according the average number of the excess packets needed for the successful decoding of all source packets while receiving in a random order the checksum packets of a given set. The average numbers of needed excess packets are plotted for all pre-computed sets:

[ xls | m2.xls | m3.xls | m4.xls | m5.xls | m6.xls | cpp ]

 

This chart shows that the performance of the random linear fountain code can be achieved also with pre-computed of various sizes.

 

For the best set of checksum packets with the size factor of 6, the probability of recovery as a function of the received packets is plotted on the chart below:

[ xls | cpp ]

 

A similar chart is given for the best pre-computed set of the size factor of 5:

[ xls | cpp ]

 

Logically, as shown in the chart below, the probability of recovery is higher when transmitting the copies from the pre-computed set of the size factor 6 than of the size factor 5:

[xls]

 

 

 

Related links:

 

-         The C++ program generating pre-computed sets (of checksum packets) of various size factors [ cpp | txt | csv ]

-         The used classes and functions [ recv.cpp | recv.h | graph.cpp | graph.h | linear.cpp | linear.h | lt.cpp | lt.h ]

-         Download all program files [zip]

-         MS Word version of this page [doc]

-         The average number of excess packets needed for successful decoding in a pre-computed set of checksum packets of constant degree [ ch | us ]

 

*   *   *