Compression is always a computationally intense project, but generic file compression has been done again and again and again and again and again (sensing a theme?). Try selecting one specific type of data and compressing that. Compression of raw tcpdump data could provide an interesting challenge. Due to the very repetitive nature of the packets headers, you could probably get quite a bit of compression out of it. Then you could compare the compression ratio of your algorithm with other conventional compression algorithms and also see how your algortim fares when combined with the more conventional algorithms.