Submerge

Experimental databasey thing

This is a very early and experimental sketch of a distributed database. Justification and explanation will be forthcoming if it ever amounts to anything. Not serious yet.

Rough picture:

• Tiered local storage
  • Small redb tables for hot data (BTrees, direct)
  • Large newel slabs for cold data (LSM, vectorized)
• Simple replication and commit protocol (clepsydra)
• Staged typed query language on top (brevet)