What we run: two nodes (one active, one standby for failover) - Each node: 2 CPU, 16 GB RAM, 500 GB storage, DNS, static IP (~$900/year per node) - Remote KMS, so the consensus key never lives on rented hardware - Target: 99% uptime (under 3.65 days of unplanned downtime per year)
What it costs:
| Item | Monthly | Annual |
|---|---|---|
| Infra (2 nodes @ $75) | $150 | $1,800 |
| Maintenance (40 hrs @ $10) | $400 | $4,800 |
| Total | $550 | $6,600 |
That's 1.57% of the ZecHub grant budget.
Pricing sources: Akash pricing calculator, DigitalOcean pricing calculator
The main validator runs on Akash Network. We always keep at least one backup node on standby for failover; that backup is not limited to Akash. If a provider goes down or censors us, the standby takes over as the validator quickly. Because the consensus key is held off-provider (see below), switching nodes doesn't mean moving or exposing the key.
Akash providers are consistently cheaper than traditional cloud VPS, sometimes by more than 50%. That keeps the budget lean while still paying for redundancy.
Measured on our live Penumbra node, restored from a pruned snapshot:
| Snapshot (height 12,598,602) | Live (height 13,604,203) | |
|---|---|---|
| App state | 21.7 GiB | 66.2 GiB |
| Block history | none | 7.6 GiB |
That works out to about 5.4 GiB/week across 9.7 weeks of chain activity. We plan around 5–10 GB/week to leave headroom.
With ~74 GiB used on a 500 GB disk, each node has room for roughly 40–80 weeks of growth at current activity, less if usage climbs. When disk gets tight, we re-sync from a fresh pruned snapshot, which resets app state to a single version. We use Rotko's snapshots for this.