- row oriented storage
- how to chunk
- Better for when we want all the values from one column
Column compression
- 데이터 패턴을 이용해 압축 Run-Length Encoding
- less data to send over the network to a server
- can keep more data stored in memory
Dictionary Compression
* 5 distinct clubs => 3bits
* apache file system parquet
* each file has metadata has its some basic information
Disadvantage
- sorted in a same way
- write to many different plcae on disk
- LSM Tree + SSTable is solution