Not sure if you have already found a satisfactory answer. If got an invite recently and feel I can provide better information than already posted.
Adding to twonuts suggestion:
Knowledge of the nature of data would be useful in pointing you to a better direction.
Look for information about information coding:
e.g. if you have long sequences of ones and zeros,
run length coding http://en.wikipedia.org/wiki/Run-length_encoding
or for more sequences of repeated data, huffman coding:http://en.wikipedia.org/wiki/Huffman_coding