A VPS for a blockchain node: disk, bandwidth and sync time
A blockchain node is a long-running service that stores a copy of the chain, exchanges blocks with other network participants over P2P and answers requests via RPC. In terms of server requirements it is closer to a busy database than to a website: it is limited by disk and bandwidth, and any downtime or lag behind the network shows up immediately. This page collects the technical criteria for choosing a VPS for test, full and validator nodes; it does not cover financial questions.
Disk: IOPS matter more than capacity
A node constantly reads and updates its state database, which means millions of small random operations. On a slow disk it can't keep up with new blocks and falls permanently behind the network, even though the CPU is almost idle. That is why project requirements almost always start with the words "NVMe SSD" rather than a core count.
- Choose only NVMe plans, and after buying, test 4K random operations with
fio: plans with a hard IOPS cap on network storage are unsuitable for nodes, even if they say SSD. - Size the disk with growth in mind: the chain gets bigger every day, and a client upgrade sometimes needs a temporary copy of the database. Plan for at least one and a half to two times the current size.
- Understand the storage modes: a pruned node keeps only recent state and takes several times less space, while an archive node keeps the full history and needs several times more. Most tasks don't need an archive.
Requirements vary widely between networks: a test node for a new project is fine with 4 cores, 8 GB of memory and a couple of hundred gigabytes, while a full node of a large network needs terabytes. A worked example for one of the heaviest networks is on the VPS for an Ethereum node page.
Bandwidth and traffic
A node sends and receives blocks to and from dozens of peers around the clock. Even without external RPC clients that is hundreds of gigabytes a month, and for popular networks several terabytes. Check not only the port speed but also the plan's traffic allowance: once it runs out, speed is often throttled to 10–100 Mbit/s and the node starts falling behind. Suitable options are in the VPS with unlimited traffic selection. The number of peers and the upload rate can be tuned in the client configuration if the link has to be shared with other services.
Sync time
Syncing from scratch, from the first block, takes anywhere from a day to several weeks on a large network and is almost entirely disk-bound. Most clients support faster modes: state synchronisation (state sync, snap sync), starting from a trusted checkpoint (checkpoint sync) or loading a ready-made database snapshot. A snapshot saves days, but download it only from the project's official sources and verify the checksum: a corrupted or tampered database won't show itself straight away.
Build the sync time into your plan: the server is billed from day one, but the node only starts doing useful work once it has caught up with the network. Daily or hourly billing is handy for checking whether the chosen plan keeps up with the network before switching to a monthly period.
Network, ports and keys
- The client's P2P port must be open for inbound connections: without them the node finds fewer peers and syncs more slowly.
- Expose the RPC port only on localhost or a private network. An open RPC gets used by other people's applications, clogging your bandwidth and disk; if outside access is needed, put a reverse proxy with authentication and rate limiting in front of it.
- Keep the clock synchronised with
chrony: consensus depends on accurate time, and a drift of seconds leads to missed slots and rejected messages. - Store validator keys encrypted and never run two copies of a node with the same key: double-signing a block is a protocol violation that the network punishes by removing the validator.
Monitoring and host rules
A node needs external monitoring: block height lag behind the network, peer count, free space and disk latency. Most clients expose metrics in Prometheus format, and alerts are best collected on a separate server; see VPS for monitoring. If several test nodes run on one VPS, they compete for the same IOPS, so make sure none of them falls behind.
And check the host's rules before paying: some providers explicitly ban blockchain nodes in their terms of service and may shut a server down based on its characteristic P2P traffic, sometimes without a refund.