Compression Opportunities. In this section we analyze the compression opportunities for the target system in the project and the provided applications. First we briefly describe the main techniques proposed so far which are the ones that are the target for the project (you can find more information about them on classical literature such as [Sal04] ): • Zero elimination: this kind of techniques compresses code or data by eliminating long strings of zeros, either by codifying them or by just removing them. From the many techniques that fall into this group those that remove zeros seemed the most interesting to us. In order to be able to eliminate part of a message without further coding, this part must be aligned (word aligned, block aligned,... depending on the particular technique). • LZn: the basic idea of this method is to use part of the input stream as a dictionary, maintaining a sliding window divided in two parts: the search buffer (left, already coded) and the look-ahead buffer (right, to be read). When a match between both buffers is found, a pointer to the occurrence in the left buffer is created and sent instead of the occurrence in the right buffer. • Table-based compression techniques: this kind of techniques creates tables with highly frequent patterns and avoids sending the whole pattern by sending the index in the table instead. There are many different techniques according to different philosophies, but they don't require compressible data to be aligned or its occurrences to be consecutive. As we have seen, there are many different compression techniques and they imply different data analysis. In order to be able to choose the compression technique that best suits our particular case of study, we have obtained a variety of data statistics not to constraint the final decision derived from applying our analysis method. We have thus used three analysis strategies to find repeating patterns: • Aligned Consecutively Repeating Patterns (ACRP). Patterns in this category are consecutive and aligned to byte size. One example is “000000010000000100000001…” where the pattern “00000001” is repeated three times (consecutive repetitions) and is byte aligned. • non-Aligned Consecutively Repeating Patterns (nACRP).. This category groups those patterns that are still consecutive but not necessarily aligned to any data size. One example is “1111000000010000000100000001…” where the pattern “00000001” is repeated three times (consecutive repetitions) and is not byte aligned. • non-Alig...
