The compression function of MySQL InnoDB is mainly used for long-term storage of cold data. For frequently accessed hot data, it may bring up to 50% performance impact, so it is rarely used in production environments. Existing compression mechanisms, such as table compression and page compression, are difficult to provide the flexibility of custom compression methods for specific fields without affecting the normal operation of the business.
Since the community MySQL InnoDB version 5.1, the table compression function has been supported. The usage method is CREATE TABLE ... ROW_FORMAT=COMPRESSED KEY_BLOCK_SIZE=8. This method requires that data must be compressed to a fixed size, such as 1K, 2K, 4K, or 8K, and once specified, it cannot be modified.
In the community MySQL InnoDB 5.7 version, page compression is kenya phone number data further supported, and the usage is CREATE TABLE ... COMPRESSION="zlib". When the page is refreshed at the Checkpoint, the page data is compressed, and when it is read to the Buffer Pool, it is decompressed. The compressed page will be kept in the Buffer Pool, with both compressed and uncompressed versions, occupying the Buffer Pool space. This compression method is related to the block size. For example, when the file system block size is 4KB, the data in a 16KB page is compressed to 9KB, and it still occupies 3 4KB file system blocks.
TaurusDB has developed and launched the field-level compression feature, which aims to provide more refined compression control for more efficient storage optimization.
2. Characteristic value
TaurusDB has launched a fine-grained field compression function, providing two compression algorithms, ZLIB and ZSTD, for VARCHAR and LOB fields. Field data is compressed when it is stored in the Page, and decompressed when the field is read to ensure that the Page size remains unchanged, and the compressed data is saved in the REDO data. Users can choose the appropriate compression algorithm and level based on actual needs, taking into account the compression ratio and the performance impact of compression/decompression operations, and compress large fields that are not frequently accessed.