relational DB: SQL interface in-memory: fast, data in mem first then -> disk distributed: scale out, multiple hosts

nodes type:

  1. aggregator node: only metadata(data about data)
  2. leaf node: data in here

1 master aggregator, 1 or more child aggregator

cluster MA A L L L

each node has SQL as interface

user query -> A -> fan out -> leaf node

DDL(data definition language) -> MA e.g. create + s/i/d/u DML(data manipulation language) -> e,g, s/i/d/u

memSQl ops on MA:

  1. UI
  2. Cli

-> MemSQL