Skip to main content

Consensus Clients

ClientVersionDateDB SizeRAMCPUNotes
Teku22.1.1Jan 2022~30 GiB~9 GiB100-300%
Lighthouse2.1.1Jan 2022~90 GiB~1.7 GiB50-200%
Nimbus1.6.0Jan 2022~40 GiB~2.3 GiB50-200%
Prysm2.1.3Jul 2022~100 GiB~4 GiB100-300%

Execution clients

For reference, here are disk, RAM and CPU requirements, as well as mainnet initial synchronization times, for different Ethereum execution clients.

Disk, RAM, CPU requirements

SSD, RAM and CPU use is after initial sync, when keeping up with head. 100% CPU is one core.

ClientVersionDateDB SizeDB GrowthRAMCPUNotes
Geth1.10.18Jun 2022~560 GiB~13.5 GiB / week8 GiB100-400%default cache size
Geth1.10.18Jun 2022~560 GiB~12 GiB / week9-10 GiB100-400%--cache 5336, max value at 16 GiB RAM, reduces DB growth rate
Geth1.10.18Jun 2022~560 GiB~8 GiB / week16-19 GiB100-400%--cache 10704, max value at 32 GiB RAM, reduces DB growth rate
Nethermind1.12.4Feb 2022~660 GiB~16 GiB / week15-16 GiB50-200%memory use w/ pruning and prune-cache 4096; 18 GiB memory and 8 cores during sync
Besu22.4.1May 2022~610 GiB~8 GiB / week8 - 9 GiB50-100%with Bonsai tries
Erigon2.28.1Oct 2022~913 GiB~18 GiB / weekSee comment50-100%Erigon will have the OS use all available RAM as a DB cache during post-sync operation, but this RAM is free to be used by other programs as needed. During sync, it may run out of memory on machines with less than 32 GiB

Notes on disk usage

  • Geth - DB size can be reduced when it grew too large by offline(!) prune
  • Nethermind - DB size can be reduced when it grow too large by online prune: Switch to full prune, manually start prune, switch back to memory prune
  • Erigon does not compress its DB, leaving that to the filesystem. With ZFS and lz4, it compresses around 1.4x. Be sure to set recordsize 16k on Erigon's dataset.

Test Systems

IOPS is random read-write IOPS measured by fio with "typical" DB parameters, 4G and 150G file, without other processes running.

Specifically fio --randrepeat=1 --ioengine=libaio --direct=1 --gtod_reduce=1 --name=test --filename=test --bs=4k --iodepth=64 --size=4G --readwrite=randrw --rwmixread=75 followed by fio --randrepeat=1 --ioengine=libaio --direct=1 --gtod_reduce=1 --name=test --filename=test --bs=4k --iodepth=64 --size=150G --readwrite=randrw --rwmixread=75, then rm test to get rid of the 150G test file. If the 150G test shows it'd take hours to complete, feel free to cut it short once the IOPS display for the test looks steady.

150G was chosen to "break through" any caching strategems the SSD uses for bursty writes. Execution clients write steadily, and the performance of an SSD under heavy write is more important than its performance with bursty writes.

Read and write latencies are measured with sudo iostat -mdx 240 2 during Geth sync, look at r_await and w_await of the second output block.

Servers have been configured with noatime and no swap to improve available IOPS.

A note on Contabo: Stability of their service is questionable.

NameRAMSSD SizeCPUr/w IOPSr/w latencyNotes
Homebrew Xeon ZFS zvol32 GiB1.2 TiBIntel Quad19.3k/6.4k (4G file) 3.5k/1k (150G file)Intel SATA SSD, 16k recordsize, stripe, xfs; fio with --bs=16k
Homebrew Xeon ZFS dataset32 GiB1.2 TiBIntel Quad38.1k/12.5k (4G file) 1.2k/500 (150G file)
Dell R420 w/ HBA32 GiB1 TBDual Intel Octo44.7k/14k (4G file) 35.9k/11k (150G file)Xeon E5-2450
Contabo L VPS SSD30 GiB800 GiBIntel Octa3.1k/1k (4G file) 2.5k/800 (150G file)This was not sufficient to sync Geth
Netcup VPS 3000 G924 GiB600 GiBAMD Hexa25.8k/8.6k (4G file) 11.2k/3.7k (150G file)2.25/6 ms
Netcup RS 8000 G9.564 GiB2 TBAMD EPYC 770247.8k/16k (4G file) 15.6k/5k (150G file)3.4/1.5 ms
OVH Baremetal NVMe32 GiB1.9 TBIntel Hexa267k/89k (4G file) 177k/59k (150G file)0.08/3.5 ms
AWS io1 w/ 10K IOPS8 GiBNAIntel Dual7.7k/2.6k (4G file) 7.6k/2.5k (150G file)t2.large, could not sync Geth. Note t2 throttles CPU
AWS gp3 w/ 16K IOPS16 GiBNAIntel Quad12.5k/4.2k (4G file) 12.2k/4.1k (150G file)m6i.xlarge

Initial sync times

NB: All execution clients need to download state after getting blocks. If state isn't "in" yet, your sync is not done. This is a heavily disk IOPS dependent operation, which is why HDD cannot be used for a node.

For Nethermind, seeing "branches" percentage reset to "0.00%" after state root changes with "Setting sync state root to" is normal and expected. With sufficient IOPS, the node will "catch up" and get in sync.

For Geth, you will see "State heal in progress" after initial sync, which will persist for a few hours if IOPS are low-ish.

This should complete in under 4 hours. If it does not, or even goes on for a week+, you do not have sufficient IOPS for Geth to "catch up" with state.

Cache size default in all tests.

ClientVersionDateTest SystemTime TakenNotes
Geth1.10.1Mar 2021Homebrew Xeon~ 10 hours
Geth1.10.9Oct 2021OVH Baremetal~ 4.5 hours
Geth1.10.9Oct 2021Netcup VPS3000~ 13 hours
Geth1.10.13Nov 2021Contabo LNeverVPS IOPS too low to finish Geth sync
Nethermind1.10.7-betaJan 2021Contabo LNeverVPS IOPS too low to finish Nethermind sync
Nethermind1.10.44Mar 2021Homebrew Xeon~ 27 hours
Nethermind1.10.9Jan 2021Netcup VPS 2000~ 20 hours
Besu22.4.1May 2022OVH Baremetal NVMe~ 30 hoursWith X_SNAP sync
Erigon2022.09.03Oct 2022EPYC 7443p~ 4 days

Getting better IOPS

Geth needs a decent amount of IOPS, as do Besu and Nethermind. Erigon can run on very low IOPS, though should also not be used with HDD.

For cloud providers, here are some results for syncing Geth.

  • AWS, gp2 or gp3 with provisioned IOPS have both been tested successfully.
  • Linode block storage, make sure to get NVMe-backed storage.
  • Netcup is sufficient as of late 2021.
  • Contabo VPS SSD cannot sync Geth as of late 2021.
  • There are reports that Digital Ocean block storage is too slow, as of late 2021.
  • Strato V-Server is too slow as of late 2021.

Dedicated servers with SSD or NVMe will always have sufficient IOPS. Do avoid hardware RAID though, see below. OVH Advance line as well as Hetzner are well-liked dedicated options; Linode or Strato or any other provider will work as well.

For own hardware, we've seen three causes of low IOPS:

  • Overheating of the SSD. Check smartctl -x. You want the SSD to be at or below 40 degrees Celsius.
  • External SSD with a USB controller that can't keep up. Samsung T5 has been shown to work, as has Samsung T7 with the right firmware. T7 is slower.
  • Hardware RAID, no TRIM support. Flash the controller to HBA and use software RAID.

In some cases, the SSD itself can't keep up, e.g. reports of this with WD Green SN350, Crucial BX500. While they sync slowly, even QLC/DRAMless SSDs can be "enough" - this depends heavily on model. Given the option, you may want to choose a "mainstream" SSD for better sync and pruning performance.