Improve the R in RDBMS: https://community.progress.com/community_groups/openedge_development/f/19/p/1704/7368.aspx#7368
(starting points for an open-source ERD tool could be http://mogwai.sourceforge.net/erdesignerng.html or http://ermaster.sourceforge.net/ )
Introduce json datatype, sharding
sequential GUIDs / clustered index wanted in a next gen database also.
What we would like to see is "cost based" index selection. The selection of the index should take into account the data that is present in the database.
The current "compiler based" index selection does not give good results in many occasions,
Concerning sequential GUIDs / clustered index this is a direct link: www.codeproject.com/.../GUIDs-as-fast-primary-keys-under-multiple-database