Infrastructure
Resource Requirements (Production)
We recommend production usage of protokit via docker and the images and services provided in the starter kit.
During our devnet testing, we observed a resource usage of at least
- 8 core CPU
- 24GB RAM
This was in a setup of 4 specialized workers and the full sovereign environment spun up.
Worker variants RAM usage:
| Worker variant | RAM usage (idle) | RAM (max) |
|---|---|---|
| Settlement Contracts (Bridging) | 2.8 GB | tbd |
| Statetransition Prover (Batch size n) | 1.8 GB | tbd |
| Transaction Prover & Runtime | 3.3 GB | tbd |
| Block Prover | 3.7 GB | 5.6 GB |
Scaling workers
For scaling workers, we currently recommend using Docker Swarm to manually scale worker variants based on usage.
Generally, if appchain usage is somewhat low, having one worker per type is enough for stable proving & settlement. Protokit however, allows unlimited horizontal scaling of workers in it’s architecture and therefore should be used if proving throughput becomes an issue.
Domain setup
Naturally, for long-running testnet deployments, you will want to host protokit behind a public domain.
Generally, there are three services that you might want to reach from the internet:
- Your app UI
- Explorer
- Grafana
In the starter-kit setup, under docker/proxy/Caddyfile you can configure the reverse-proxy’s (Caddy) mappings.
Change the directives to your domain and sub-domain respectively.
Since both your app UI and the explorer will (most likely) be built using next.js, you’ll have to host them under separate subdomains, sub-paths won’t work.
I.e. protokit.dev -> UI and explorer.protokit.dev -> Explorer