My answer (above) works as follows:

For every length you want to consider
  For each packet
    for each substring of length N
      increment count of N's occurrence by 1
Look at all the data you've amassed about which substrings
occur with what frequency and spit out some data.
[download]

Unless you specify what your "length vs frequency" preference is you can not get a generic answer. If what you are looking for is "what are the largest, most common substrings that occur in more than %90 of the packets" then you can do it.