If the size is before compression, then after compression, the strip size stored on disk will be not uniform which doesn't look good. But if it's after compression, then how did hive know the size is 250M after compression? Will hive compress some, check whether it reaches 250M, if not reached, then add more and compress, repeat over again and again until it reaches 250M. But this looks like not cost effective. Anyone could help me understand?
