Key considerations for KV store

HA - Scalability

• Sharding -> without knowledge of the kind of key it is better to shard by hash of key -> consistent hashing
• Replication
  • Need to implement primary/replica pattern for shards
  • Index always to primary, replicate to replicas
  • Mark as ready when all have applied a change
  • If primary shard is down, promote a replica
• Need of a leader to keep cluster state
• Routing/allocation algorithms

Durability

• Different layers/tiers of storage (cached/disk)

Reliability

• Error handling
  • Handle node failure
  • Handle shard failure
  • Handle write failure
  • Route to healthy node/shard

Consistency

• Transactions ?
• Entry versioning ?

Goals for this simple implementation :

• Represent nodes as separate in memory structs
• Single table approach
• In memory map used as backing store
• Main goroutine to act as master and track cluster state
• Only main node to receive requests
• Channels as node-node communication
• Constant number of nodes=5 and shards=10 (1- primary and 10 replicas)
• Handle some errors
  • Node unavailable ->
    • reroute request to replica
    • turn replica into primary
    • Create a new replica in another node

How to run

1. Clone this repo
2. Install docker: https://docs.docker.com/get-docker/
3. Build the binaries by running: make build-server && make build-client
4. Run the server with: docker run -p 1338:1338 kv-store-server:latest
5. Run the client with:

docker run --network="host" kv-store-client:latest localhost:1338 put name luis
docker run --network="host" kv-store-client:latest localhost:1338 get name