Groupcache DB Experiment - Revisited 2019

This is the second revision of Capotej's Groupcache DB Experiment.

I decided to replay again his experiment using newer techniques, such the use of go mod, protobuf and error wrapping.

This project simulates a scenario wherein a few frontends running groupcache are fronting a slow database. See his blog post about it for more details.

Getting it running

The following commands will set up this topology:

Build everything

1. git clone https://github.com/cirello-io/groupcache-experiment.git
2. go build ./cmd/backend
3. go build ./cmd/frontend
4. go build ./cmd/cli

Start DB server

1. ./backend

This starts a deliberately slow k/v datastore on :8080

Start Multiple Frontends

1. ./frontend -listen "http://localhost:8001" -frontend "localhost:9001"
2. ./frontend -listen "http://localhost:8002" -frontend "localhost:9002"
3. ./frontend -listen "http://localhost:8003" -frontend "localhost:9003"

Use the CLI to set/get values

1. ./cli -set -k foo -v bar
2. ./cli -get -k foo should see bar in 300 ms
3. ./cli -get -k foo should see bar instantly