Partially-Decompressible Dictionary Based Compression Format for All Flash Array

The demand for managing a huge amount of data, so-called “big data”, has been increasing in all kinds of fields. All Flash Array (AFA) is a possible solution to face this difficulty, since it provides high-bandwidth and low latency data management with several SSDs and computational resources. The most troublesome problem of exploiting AFA is that each SSD has upper limit of writing times, which shortens the lifespan of the flash memory cells. Data compression techniques can reduce the amount of data to write and solve the lifetime problem. In order to improve the compression ratio, data blocks should be compressed in bulk to leverage inter-block common features. However, partial read requests are problematic because existing compression methods do not support the partial decompression feature. Without this feature, storage systems are required to decompress entire compressed blocks to read any fraction of data. We propose a dictionary coding based compression technique, which can leverage the benefit of the bulk compression method and support the partial decompression feature. This characteristic provides competitive decompression speed with the case that each block is individually compressed.