This is the task for the schema change project documented on the wiki at [[https://www.mediawiki.org/wiki/User:Brion_VIBBER/Compacting_the_revision_table_round_2|User:Brion VIBBER/Compacting the revision table round 2]]. Part of the description is copied below.
Per ongoing discussion in ArchCom and at WikiDev17 about performance, future requirements, and future-proofing for table size it's proposed to do a major overhaul of the revision table, combining the following improvements:
* Normalization of frequently duplicated data to separate tables, reducing the dupe strings to integer keys
* Separation of content-specific from general-revision metadata to support:
* Multi-content revisions allowing for storing of multiple content blobs per revision -- not related to compaction, but of great interest for structured data additions planned for multimedia and articles
* general reduction in revision table width / on-disk size will make schema changes easier in future
* trying to avoid inconsistencies in live index deployments
* ideally all indexes should fit on all servers, making it easier to switch database backend around in production
Wikimedia production tasks:
* {T166733}
* {T188327}
* {T215466}