Conversation

Implemented a crazy idea, and it successfully reduced the amount of WAL to one-third in the pgbench benchmark.

Replying to

Wow, really? Will we hear more in a while?

Replying to

Yes, just compressed the full-page-writes in WAL by pglz.

Replying to

sounds very interesting!

Replying to

I am also working on something. A feature similar to MySQL's INSERT IGNORE. Close to publishing.

Replying to

INSERT CORRUPT INTO? ;-)

Replying to

Motivated in part by getting general upsert infrastructure in place.