Skip to main content

Resource Matrix

These are starting envelopes for production planning. Validate them with chain sync lag, RPC p95 latency, disk growth, compaction pressure, and peer count before committing an SLO.

ChainNode typeCPURAMDiskPrimary scale signal
EthereumFull execution + consensus4-8 vCPU16-32 GiB2+ TiB NVMeHead lag, peer count, RPC latency
EthereumArchive / trace16+ vCPU64+ GiBMany TiB NVMeTrace latency, compaction stalls, free space
SuiFull node4-8 vCPU16-32 GiBFast SSD sized for checkpoint growthCheckpoint lag, RocksDB pressure, API latency
SuiArchive / long-retention API8-16 vCPU32-64 GiBMulti-TiB SSD/NVMeHistorical query latency, checkpoint retention growth
AptosFullnode4 vCPU16 GiB2+ TiB SSDREST p95, sync lag, storage growth
AptosArchive / indexer-backed API8+ vCPU plus DB capacity32+ GiBMulti-TiB SSD plus indexer databaseProcessor lag, GraphQL latency, database saturation
SolanaFull RPC node16+ vCPU128+ GiBFast local NVMeSlot lag, account index memory, RPC p95
SolanaArchive / historical RPC32+ vCPU256+ GiBLarge NVMe set plus ledger retention capacityLedger retention, blockstore growth, replay latency
tip

Separate heavy indexing, tracing, and public RPC traffic from validator-critical or consensus-adjacent workloads. Saturating disk or CPU on a node process usually shows up as sync lag before it shows up as a clean crash.