Hardware Requirements

Minimum Specs

Fleet Size	vCPU	RAM	Disk
< 10 agents	2	4 GB	40 GB SSD
10–50 agents	2	8 GB	80 GB SSD
50–500 agents	4	16 GB	200 GB
500–2000 agents	8	32 GB	500 GB
> 2000 agents	multiple nodes (postgres replica + horizontal hub-api)

Consumption per Agent

• Ingest: ~1 KB / 15 sec / agent → 50 agents ≈ 17 MB/u
• Inventory: ~80 KB / 6u / agent
• Telemetry 30 days: ~50 MB / agent in TimescaleDB (after compression)

Backup Strategy

# Daily logical dump
pg_dump -U monsys -h postgres monsys | zstd -19 > /backups/$(date -I).sql.zst

# 30 day retention
find /backups -name "*.sql.zst" -mtime +30 -delete

For production: also create a hot replica via streaming replication with pg_basebackup on a second node.

Scaling Advice

• Split hub-api and hub-ingest into separate instances when > 200 agents.
• Postgres connection pool: minimum max_connections = 100; use PgBouncer for > 500 agents.
• TimescaleDB compression after 7 days — saves 80–95% disk space.
• For multi-region: cross-region replication is not needed — agents choose the closest ingest endpoint via DNS.