Db2 - simple, local database

link to build

link to run

link to commands

link to Things to Tackle

build

• server - make server
• cli - make cli

run

server

./bin/db2.out

client

./bin/cli.out [command] {arg1 .. argk}

commands

key/value

• get [key]
• set [key] [value]
• delete [key]
• filter []

timeseries

• create [name]
• add [series_descriptor] [value]
• get_range [start] [end]
• start [series_descriptor]
• end [series_descriptor]

Things to Tackle

• memory - currently mempool is a wrapper around malloc and free.
  • consider kv memory, timeseries, and the future (tables, etc.)
  • the idea is to not have to use syscalls for memory
• query on kv -
  • how to store and handle filter results?
  • how to provide easy use, extendability and not too much work?
  • pipe queries ? meaning support this - filter key startswith 'user' && value.email !has '@'
  • so that the second part only looks at the results of the first, not the whole kv
  • when value is an object, how to access its fields for querying?
• cache
  • what it takes to turn kv into a cache
• config
  • only what must be decided before build
  • total memory
  • max kv entries (optional, if you don't intend to use kv)
  • limit kv key / value size
  • max number of time-series
  • max entries per time-series
  • if and how to compress time-series
  • cache defualt ttl
• management cli -
  • add / remove services like - ./db2_mgmt start kv
  • check status for services, memory etc.
  • flag / block connections
  • add / remove nodes (far future)
• time-series compression
  • ask Mr. GPT - "what are some popular strategies to compress a time-series?"
  • implement 1 or 2, make clear interface to add more
  • how to store and access these compressed series?
• ingest in bulk
• support http/s connections