
web3 infrastructure
Validator vs RPC vs Full Node: Which One You Need
This guide breaks down the validator, full, RPC, archive and light node roles, with 2026 hardware figures and a framework for choosing between in-house operation and managed infrastructure.
SEP 24, 2026
Last updated SEP 24, 2026 · V1
TL;DR
- A validator node takes part in consensus, is eligible for protocol rewards and carries slashing exposure.
- A full node verifies every block independently and receives no protocol rewards.
- An RPC node answers application requests and is tuned for throughput and latency.
- An archive node keeps every historical state for explorers, analytics and audits.
- A sequencer orders rollup transactions and is typically run by one operator.
- Most institutions need verification and data access before they need consensus participation.
Institutions entering proof-of-stake networks often start by asking which node to run, but the answer depends on whether they need to verify chain data, serve applications or take part in consensus.
This guide breaks down the validator, full, RPC, archive and light node roles, with 2026 hardware figures and a framework for choosing between in-house operation and managed infrastructure.
The Five Node Roles
The five node types differ in consensus duties, reward eligibility, storage and the teams that use them.
| Node type | Runs consensus | Protocol rewards | Disk actually used | Recommended hardware | Who needs it |
|---|---|---|---|---|---|
| Validator | Yes | Yes, with slashing exposure | Same as a full node (execution + consensus client); the validator client adds negligible disk | 4 TB NVMe, 64 GB RAM, 8 cores / 16 threads | Staking providers, custodians and exchanges that offer staking |
| Full node | No | No | ~1.3 TB on Geth before history pruning; ~420 GB on Erigon v3.5 | 4 TB NVMe, 32 GB RAM, 4 cores / 8 threads | Any team that verifies its own chain data |
| RPC node | No | No | Full-node or archive footprint, depending on the query range served | Full-node baseline at minimum; scale RAM, CPU and replicas to request load | Wallets, exchanges, dApps |
| Archive node | No | No | ~2 TB on Geth path-based (~6.5 TB with historical trie data); 2.03 TB on Erigon; ~18–20 TB on legacy hash-based Geth | A single 4 TB NVMe fits path-based Geth or Erigon; hash-based Geth needs large NVMe arrays | Explorers, analytics, compliance teams |
| Light node | No | No | Headers and sync-committee updates only; negligible | Consumer device | Mobile and embedded clients |
Hardware figures come from 2026 operator guidance from Cherry Servers, Chainstack and the Geth documentation. Actual sizing varies by client, pruning settings and hosting model. Read more about why Everstake selected Cherry Servers for running nodes on Solana.
For Solana, colocation runs $300–500 per month in rack and power after a $20,000–50,000 hardware purchase.
| Node type | CPU | RAM | Disk |
|---|---|---|---|
| Validator | 12 cores / 24 threads or more | 256 GB or more | Separate NVMe drives: accounts 1 TB+, ledger 1 TB+, snapshots 500 GB+ |
| RPC node | 16 cores / 32 threads or more | 512 GB or more (for all account indexes) | Same as validator; accounts and ledger must be on separate disks |
Cloud hosting adds $900–1,200 monthly in egress fees for a Solana validator, according to the Everstake analysis of Solana validator hardware vs cloud.
What a Validator Node Does
A validator node proposes blocks, attests to blocks from other validators and signs each message with a staked key. The protocol rewards correct duties and penalizes downtime or conflicting signatures.
On Ethereum, a validator runs three software components:
- an execution client such as Geth, Nethermind or Erigon,
- a consensus client such as Lighthouse or Prysm,
- a validator client that holds the signing key.
Beyond hardware, validator node requirements obligate the operator to:
- protect signing keys from theft and from use on two machines at once,
- keep the node online, since missed duties reduce rewards,
- apply client upgrades before each network hard fork,
- monitor for double-signing risk during failover.
The Pectra upgrade on May 7, 2025 raised the maximum effective balance to 2,048 ETH per validator. It also cut the initial slashing penalty from 1/32 to 1/4,096 of effective balance.
A correlation penalty grows with the share of stake slashed in the same period. A slashed validator is forced to exit, and withdrawals are delayed by about 36 days, or 8,192 epochs, according to Chorus One.
The minimum to activate an Ethereum validator remains 32 ETH. A dedicated validator for institutions gives one organization its own validator keys and stake, isolated from pooled deposits.
Full Nodes: Verification Without Consensus
A full node downloads every block and re-executes every transaction against protocol rules. It holds no stake, signs no consensus messages and receives no protocol rewards.
An exchange that credits deposits from a third-party API trusts that provider’s view of the chain. A full node lets the exchange confirm balances and finality against its own copy of the state.
Independent verification supports:
- deposit and withdrawal reconciliation,
- proof-of-reserves checks,
- audit trails built from the institution’s own chain data,
- fallback access during an outage at an external provider.
A light node downloads block headers and requests proofs from full nodes. It suits mobile wallets, while a full node suits systems that must verify every transaction.
RPC and Archive Nodes: Serving Application Data
An RPC node exposes the chain through an API, usually JSON-RPC, so wallets and backends can read state and send transactions. Operators tune it for request throughput and response latency.
In an RPC node vs validator node comparison, the validator must reach other validators fast enough to vote on time. The RPC node must handle large volumes of concurrent requests.
On Solana, it is recommended to have 256 GB RAM minimum for validators and 512–768 GB for RPC nodes.
An RPC node running a full client verifies each block it receives. It casts no votes and proposes no blocks, so it plays no part in consensus.
An archive node keeps every historical state, so it can answer balance queries at any past block. Common users include:
- block explorers,
- tax and accounting tools,
- analytics platforms,
- forensic and compliance teams.
A full node prunes old state and keeps recent data. A path-based Geth archive node needs about 2 TB on Ethereum mainnet, or 6.5 TB with historical trie data.
Erigon archive nodes need 2.03 TB, full 419.04 GB, minimal 378.97 GB, measured on v3.5.2, according to Erigon’s Github in March 2026. Legacy hash-based Geth archives reach 18–20 TB.
A full Solana archive exceeds 400 TB and grows by several terabytes a month, according to Cherry Servers. Partial archives that keep recent weeks or months start at 10-50 TB.
Everstake runs SWQoS RPC access on Solana via Blockspace, which routes transactions by stake weight.
Sequencers and Why They Are Not Validators
A sequencer receives rollup transactions, orders them, executes them and posts batches to a base chain. It casts no staked consensus vote, because the base chain settles the result.
In 2026, Arbitrum, Base and OP Mainnet each run a sequencer operated by one entity. Users can bypass a censoring sequencer through forced inclusion on L1, which is slower and costs more.
Stablecoin chains restrict block production to selected operators, as rollups do with sequencers, yet those operators still run BFT consensus. Arc, launched by Circle on September 16, 2026, runs Malachite, a Tendermint BFT engine, over a proof-of-authority validator set selected by Circle.
Tempo, incubated by Stripe and Paradigm, went live on March 18, 2026 with Simplex BFT consensus and permissioned validators secured without staking. The Everstake Tempo explainer notes a stated roadmap toward permissionless validation.
Which Role Fits Which Institution
The right mix depends on whether the institution stakes client assets, holds custody or moves funds. Hardware and uptime targets also depend on how a network handles institutional workloads.
| Institution | Full node | RPC node | Archive node | Validator |
| Exchange | Required | Required | Optional | When offering staking |
| Custodian | Required | Required | Optional | When offering staking |
| Asset manager | Recommended | Recommended | Recommended | Required for staked products |
| Payments company | Required | Required | Optional | By invitation on Arc and Tempo |
Exchange
An exchange needs full nodes to confirm deposits and RPC capacity for withdrawals during traffic spikes. Validators become relevant when it offers staking to clients.
Custodian
A custodian needs full nodes to verify the balances it reports to clients and auditors. Staking custodians add validators, with signing keys held in HSMs or an MPC system.
Asset Manager
An asset manager running a staked ETP needs validator operations and reward reporting. Archive access supports NAV checks and historical reward reconciliation.
Payments Company
A payments company needs low-latency RPC and full nodes on each chain it settles on. On Arc and Tempo, validator seats are assigned by the network, so firms outside that set connect through RPC.
Run It or Buy It
Operating nodes in-house means owning every incident, upgrade and key procedure. The decision rests on staffing, key control and time to launch.
In-house operation requires:
- a 24/7 on-call rotation with defined escalation paths,
- client upgrades tested on testnet before each hard fork,
- key ceremonies with documented custody of signing keys,
- failover that never runs two active signers for one key,
- monitoring for missed attestations, sync lag and disk growth.
| Factor | In-house | Everstake VaaS |
| Key control | Held by the institution | Non-custodial; client retains asset control |
| On-call rotation | Internal team, 24/7 | Everstake team, 24/7 |
| Client upgrades | Internal | Everstake |
| Slashing protection | Self-managed | Technical mitigation included |
| Time to launch | Depends on hiring and procurement | 72 hours (or custom) |
| Security attestations | Own audits | SOC 2 Type II, ISO/IEC 27001 |
Everstake VaaS deploys dedicated validator infrastructure in 72 hours. Everstake reports the following certifications, assessments and compliance frameworks:
- SOC 2 Type II,
- ISO/IEC 27001,
- NIST CSF,
- ITGC,
- GDPR,
- CCPA.
Teams that want their brand on the validator can use Everstake white-label validators. Regulated entities require a security review before deploying staking infrastructure, according to the Everstake institutional staking integration guide.
FAQ
What is the difference between a full node and a validator?
A full node verifies every block and holds no stake. A validator also votes in consensus with at least 32 ETH on Ethereum, which exposes it to slashing.
Everstake operates validators for institutions that want consensus participation without running the stack.
Do I need to run my own node?
Not always. Teams that verify deposits or report balances benefit from at least 1 full node under their control.
Can an RPC node validate transactions?
Yes, an RPC node running a full client checks each block against protocol rules. It casts 0 consensus votes and receives no protocol rewards.
What does a sequencer do?
A sequencer orders rollup transactions, executes them and posts batches to a base chain. Arbitrum, Base and OP Mainnet each rely on 1 sequencer operator in 2026. An Everstake validator, by contrast, votes in consensus rounds alongside other operators.
What is the difference between a full node and an archive node?
A full node keeps recent state and prunes older history. An Ethereum archive node needs about 2 TB on path-based Geth and 2.03 TB on Erigon.
What is the difference between a light node and a full node?
A light node downloads block headers and relies on full nodes for proofs. A full node re-executes every transaction and needs 4–8 TB of NVMe storage on Ethereum in 2026.
Do full nodes receive rewards?
No, full nodes receive 0 protocol rewards because they hold no stake. RPC operators charge for API access.
Share with your network