Hacker News new | past | comments | ask | show | jobs | submit login

Doing compression on the ZFS level is significantly worse than InnoDB compression.

InnoDB has a lot of really smart optimizations which make it much better than just zipping things up. Included are the modification log (so you only have to re-compress occasionally) and a dynamic scaling ability to keep compressed pages in memory rather than always decompressing. These optimizations are really only possible with an understanding of the data.

I would only consider ZFS for something like an append-only data warehouse type system.




I'm sure you would want the compression done by the database and not the filesystem, since there are many ways to do compression to fit specific applications, and the database knows what it's trying to do. I read a little bit of the MySQL docs regarding how it uses compression, and it sounded pretty different than general-purpose compression.


Do you have any references?


This is a good intro for those (like me) who are unfamiliar:

https://blogs.oracle.com/mysqlinnodb/entry/innodb_compressio...

I know that a lot of work has been done on InnoDB compression, but I didn't quite grasp the extent, thinking it wasn't far different from Postgres ( http://www.postgresql.org/docs/current/static/storage-toast.... ).




Consider applying for YC's Spring batch! Applications are open till Feb 11.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: