This is the full report of the run “Fleet sizing: e2 baseline against c3 (A vs B)”. Back to the run page.
r20260923-0708 · Fleet A · e2-custom-4-8192

Telcoin Network benchmark: mixed workload

Throughput is measured from blocks the chain produced (own senders vs every sender), resource use from Cloud Monitoring, Prometheus and the on-node sampler. Summaries cover the mixed phase. Hover any number for the JSON field it came from.

10 blaster(s), 6K senders, chain 32286, schema tn-blast-combined/v2 · mixed phase 2026-09-23 08:15–08:55 UTC · 1 observers, 10 validators

Preflight: WTEL from genesis (0x0000…37e1), v3 fee tier 3000, v4 available, Multicall3 present · Cooldown ended: idle_30s ×10

Headline numbers

Sustained mixed TPS (own)
3,569.4 tx/s
best 60-s mean of our own landed tx in the mixed phase
Sustained mixed TPS (chain)
3,569.4 tx/s
all tx in blocks, same window
Landed / accepted
99.6 %
7.35M of 7.38M accepted
Reverted (own)
0
0.00 % of landed
Latency p50
7,595 ms
submit → block timestamp; seen p50 22,772 ms
Latency p95
42,041 ms
seen p95 61,340 ms
Gas / s (own)
113M gas/s
chain 113M gas/s
Capacity at mix
4,268.6 tx/s
gas-limit basis 2,733.5 tx/s; block gas limit 30M

Per transaction type

Landed counts only our own senders; latency is submit → block timestamp.

TypeSubmittedAcceptedLandedRevertedUnlandedSkipsGas limitGas used avgp50p95
native_transfer2.19M2.19M2.18M014.5K024.1K21K7,573 ms41,447 ms
batched_self_send1.1M1.1M1.1M04.01K024.1K21K7,708 ms43,372 ms
uniswap_v2_swap964K964K960K03.47K86180K106K7,595 ms42,182 ms
stablecoin_transfer831K831K828K02.85K375K43.9K7,584 ms41,681 ms
wtel_wrap663K663K660K02.38K050K27.9K7,601 ms41,661 ms
wtel_unwrap438K438K436K01.51K2545K35.2K7,576 ms41,698 ms
stablecoin_approve331K331K330K01.19K060K31.8K7,602 ms41,501 ms
uniswap_v2_add_liquidity215K215K214K077986250K128K7,584 ms41,929 ms
contract_deploy165K165K165K05510140K74.2K7,592 ms41,965 ms
uniswap_v4_swap161K161K160K0578114250K124K7,561 ms42,163 ms
faucet_drip109K109K109K0184168K150K107K6,942 ms44,064 ms
uniswap_v3_swap108K108K107K040188220K120K7,564 ms41,079 ms
uniswap_v2_remove_liquidity106K106K105K0416156220K129K7,618 ms42,332 ms

Submission by RPC node

NodeSubmittedAcceptedAccepted %Top errorsSend p50Send p95
bench-validator-01
bench-validator-01:8545
904K890K98.5 %known 8.7K10 ms35 ms
bench-validator-02
bench-validator-02:8545
1.06M1.05M98.2 %known 10.2K10 ms38 ms
bench-validator-03
bench-validator-03:8545
16.2K16.2K100.0 %none41 ms147 ms
bench-validator-04
bench-validator-04:8545
387K367K94.7 %known 11K11 ms53 ms
bench-validator-05
bench-validator-05:8545
580K560K96.5 %known 7.79K13 ms46 ms
bench-validator-06
bench-validator-06:8545
736K716K97.4 %known 11K14 ms42 ms
bench-validator-07
bench-validator-07:8545
1.02M1.02M99.9 %known 31910 ms36 ms
bench-validator-08
bench-validator-08:8545
1.18M1.18M100.0 %known 7611 ms38 ms
bench-validator-09
bench-validator-09:8545
948K931K98.3 %known 11.5K10 ms34 ms
bench-validator-10
bench-validator-10:8545
738K716K97.1 %known 12.4K10 ms30 ms

Throughput over time

Shaded: phases outside the mixed window. Hairlines: epoch closes. Dashed orange: chaos events.

Landed transactions per second

own senders vs chain total, per block-timestamp second

Gas per second

own vs chain gas used, per block-timestamp second

CPU and memory per node

CPU utilization

compute.googleapis.com/instance/cpu/utilization, 60-s mean, % of all vCPUs; summaries over the mixed phase

Table view
NodeClassMeanp95Peak 60-s meanMax
bench-observer-11observer23.0 %27.3 %28.0 %28.0 %
bench-validator-01validator44.6 %48.8 %49.3 %49.3 %
bench-validator-02validator42.4 %47.7 %52.3 %52.3 %
bench-validator-03validator51.4 %55.1 %58.7 %58.7 %
bench-validator-04validator47.8 %54.7 %59.1 %59.1 %
bench-validator-05validator35.8 %45.2 %49.8 %49.8 %
bench-validator-06validator43.1 %49.3 %52.9 %52.9 %
bench-validator-07validator47.3 %52.5 %53.8 %53.8 %
bench-validator-08validator47.9 %52.2 %54.4 %54.4 %
bench-validator-09validator43.1 %46.6 %47.6 %47.6 %
bench-validator-10validator45.2 %48.8 %50.6 %50.6 %

CPU: mean vs peak 60-s mean

One row per node, over the mixed phase.

telcoin resident memory

agent processes/rss_usage summed over telcoin pids; summaries over the mixed phase

Table view
NodeClassMeanp95Peak 60-s meanMax
bench-observer-11observer2.02 GiB3.32 GiB3.67 GiB3.67 GiB
bench-validator-01validator4.74 GiB6.37 GiB6.45 GiB6.45 GiB
bench-validator-02validator4.81 GiB6.52 GiB6.57 GiB6.57 GiB
bench-validator-03validator4.96 GiB6.63 GiB6.66 GiB6.66 GiB
bench-validator-04validator4.98 GiB6.66 GiB6.68 GiB6.68 GiB
bench-validator-05validator4.22 GiB6.26 GiB6.43 GiB6.43 GiB
bench-validator-06validator5.02 GiB6.54 GiB6.58 GiB6.58 GiB
bench-validator-07validator4.77 GiB6.54 GiB6.59 GiB6.59 GiB
bench-validator-08validator4.65 GiB6.44 GiB6.51 GiB6.51 GiB
bench-validator-09validator5.00 GiB6.61 GiB6.63 GiB6.63 GiB
bench-validator-10validator5.07 GiB6.66 GiB6.68 GiB6.68 GiB

telcoin RSS: mean vs peak 60-s mean

One row per node, over the mixed phase.

Memory used

agent memory/percent_used, state=used; summaries over the mixed phase

Table view
NodeClassMeanp95Peak 60-s meanMax
bench-observer-11observer24.7 %37.0 %38.1 %38.1 %
bench-validator-01validator54.2 %72.6 %72.9 %72.9 %
bench-validator-02validator55.5 %76.4 %78.6 %78.6 %
bench-validator-03validator57.5 %77.8 %78.6 %78.6 %
bench-validator-04validator57.6 %80.3 %81.8 %81.8 %
bench-validator-05validator43.7 %74.2 %76.3 %76.3 %
bench-validator-06validator57.4 %75.4 %76.5 %76.5 %
bench-validator-07validator55.0 %78.2 %80.0 %80.0 %
bench-validator-08validator52.8 %72.0 %74.3 %74.3 %
bench-validator-09validator56.9 %74.1 %74.9 %74.9 %
bench-validator-10validator58.3 %77.8 %78.1 %78.1 %

Memory used: mean vs peak 60-s mean

One row per node, over the mixed phase.

Disk

Disk throttling (validator)
not collected –
metric type unavailable in project
Disk performance degraded (validator)
none –
any sample

Disk read + write throughput

instance/disk/{read,write}_bytes_count rate, all devices; summaries over the mixed phase

Table view
NodeClassMeanp95Peak 60-s meanMaxLimit
bench-observer-11observer39.6 MiB/s60.5 MiB/s61.2 MiB/s61.2 MiB/s240.0 MiB/s
bench-validator-01validator51.9 MiB/s71.7 MiB/s75.0 MiB/s75.0 MiB/s240.0 MiB/s
bench-validator-02validator58.0 MiB/s73.3 MiB/s75.5 MiB/s75.5 MiB/s240.0 MiB/s
bench-validator-03validator82.4 MiB/s102.5 MiB/s104.0 MiB/s104.0 MiB/s240.0 MiB/s
bench-validator-04validator96.1 MiB/s114.6 MiB/s120.4 MiB/s120.4 MiB/s240.0 MiB/s
bench-validator-05validator69.4 MiB/s121.7 MiB/s124.4 MiB/s124.4 MiB/s240.0 MiB/s
bench-validator-06validator54.6 MiB/s71.9 MiB/s75.0 MiB/s75.0 MiB/s240.0 MiB/s
bench-validator-07validator56.0 MiB/s74.6 MiB/s75.7 MiB/s75.7 MiB/s240.0 MiB/s
bench-validator-08validator52.9 MiB/s73.3 MiB/s75.8 MiB/s75.8 MiB/s240.0 MiB/s
bench-validator-09validator100.1 MiB/s121.8 MiB/s125.1 MiB/s125.1 MiB/s240.0 MiB/s
bench-validator-10validator96.7 MiB/s123.1 MiB/s126.3 MiB/s126.3 MiB/s240.0 MiB/s

Disk read + write IOPS

instance/disk/{read,write}_ops_count rate, all devices; summaries over the mixed phase

Table view
NodeClassMeanp95Peak 60-s meanMaxLimit
bench-observer-11observer3.02K IOPS3.59K IOPS3.71K IOPS3.71K IOPS9K IOPS
bench-validator-01validator3.15K IOPS3.63K IOPS4.26K IOPS4.26K IOPS9K IOPS
bench-validator-02validator3.12K IOPS3.84K IOPS4.97K IOPS4.97K IOPS9K IOPS
bench-validator-03validator2.1K IOPS4.04K IOPS4.26K IOPS4.26K IOPS9K IOPS
bench-validator-04validator2.32K IOPS4.48K IOPS4.95K IOPS4.95K IOPS9K IOPS
bench-validator-05validator1.96K IOPS3.84K IOPS4.34K IOPS4.34K IOPS9K IOPS
bench-validator-06validator3.19K IOPS3.86K IOPS4.23K IOPS4.23K IOPS9K IOPS
bench-validator-07validator3.28K IOPS4.23K IOPS5.72K IOPS5.72K IOPS9K IOPS
bench-validator-08validator3.1K IOPS3.44K IOPS3.95K IOPS3.95K IOPS9K IOPS
bench-validator-09validator1.93K IOPS3.13K IOPS3.3K IOPS3.3K IOPS9K IOPS
bench-validator-10validator2.33K IOPS4.7K IOPS5.02K IOPS5.02K IOPS9K IOPS
Disk limits used
NodeMachinevCPUDiskIOPS limitMiB/s limitBasis
bench-observer-11e2-custom-4-81924pd-ssd 100 GB9K IOPS240.0 MiB/spd-ssd 100 GB: 6,000 + 30/GB = 9,000 IOPS, 240 + 0.48/GB = 288 MiB/s; e2 4 vCPU cap 15,000 IOPS / 240 MiB/s
bench-validator-01e2-custom-4-81924pd-ssd 100 GB9K IOPS240.0 MiB/spd-ssd 100 GB: 6,000 + 30/GB = 9,000 IOPS, 240 + 0.48/GB = 288 MiB/s; e2 4 vCPU cap 15,000 IOPS / 240 MiB/s
bench-validator-02e2-custom-4-81924pd-ssd 100 GB9K IOPS240.0 MiB/spd-ssd 100 GB: 6,000 + 30/GB = 9,000 IOPS, 240 + 0.48/GB = 288 MiB/s; e2 4 vCPU cap 15,000 IOPS / 240 MiB/s
bench-validator-03e2-custom-4-81924pd-ssd 100 GB9K IOPS240.0 MiB/spd-ssd 100 GB: 6,000 + 30/GB = 9,000 IOPS, 240 + 0.48/GB = 288 MiB/s; e2 4 vCPU cap 15,000 IOPS / 240 MiB/s
bench-validator-04e2-custom-4-81924pd-ssd 100 GB9K IOPS240.0 MiB/spd-ssd 100 GB: 6,000 + 30/GB = 9,000 IOPS, 240 + 0.48/GB = 288 MiB/s; e2 4 vCPU cap 15,000 IOPS / 240 MiB/s
bench-validator-05e2-custom-4-81924pd-ssd 100 GB9K IOPS240.0 MiB/spd-ssd 100 GB: 6,000 + 30/GB = 9,000 IOPS, 240 + 0.48/GB = 288 MiB/s; e2 4 vCPU cap 15,000 IOPS / 240 MiB/s
bench-validator-06e2-custom-4-81924pd-ssd 100 GB9K IOPS240.0 MiB/spd-ssd 100 GB: 6,000 + 30/GB = 9,000 IOPS, 240 + 0.48/GB = 288 MiB/s; e2 4 vCPU cap 15,000 IOPS / 240 MiB/s
bench-validator-07e2-custom-4-81924pd-ssd 100 GB9K IOPS240.0 MiB/spd-ssd 100 GB: 6,000 + 30/GB = 9,000 IOPS, 240 + 0.48/GB = 288 MiB/s; e2 4 vCPU cap 15,000 IOPS / 240 MiB/s
bench-validator-08e2-custom-4-81924pd-ssd 100 GB9K IOPS240.0 MiB/spd-ssd 100 GB: 6,000 + 30/GB = 9,000 IOPS, 240 + 0.48/GB = 288 MiB/s; e2 4 vCPU cap 15,000 IOPS / 240 MiB/s
bench-validator-09e2-custom-4-81924pd-ssd 100 GB9K IOPS240.0 MiB/spd-ssd 100 GB: 6,000 + 30/GB = 9,000 IOPS, 240 + 0.48/GB = 288 MiB/s; e2 4 vCPU cap 15,000 IOPS / 240 MiB/s
bench-validator-10e2-custom-4-81924pd-ssd 100 GB9K IOPS240.0 MiB/spd-ssd 100 GB: 6,000 + 30/GB = 9,000 IOPS, 240 + 0.48/GB = 288 MiB/s; e2 4 vCPU cap 15,000 IOPS / 240 MiB/s

Network

Network sent

instance/network/sent_bytes_count rate; summaries over the mixed phase

Table view
NodeClassMeanp95Peak 60-s meanMax
bench-observer-11observer0.5 Mbps0.7 Mbps0.8 Mbps0.8 Mbps
bench-validator-01validator39.2 Mbps43.1 Mbps43.9 Mbps43.9 Mbps
bench-validator-02validator41.0 Mbps45.7 Mbps55.3 Mbps55.3 Mbps
bench-validator-03validator39.8 Mbps45.6 Mbps46.4 Mbps46.4 Mbps
bench-validator-04validator37.3 Mbps44.6 Mbps45.2 Mbps45.2 Mbps
bench-validator-05validator34.4 Mbps53.8 Mbps55.5 Mbps55.5 Mbps
bench-validator-06validator43.7 Mbps55.7 Mbps60.6 Mbps60.6 Mbps
bench-validator-07validator40.8 Mbps45.5 Mbps46.6 Mbps46.6 Mbps
bench-validator-08validator42.7 Mbps48.1 Mbps49.8 Mbps49.8 Mbps
bench-validator-09validator40.5 Mbps44.6 Mbps47.7 Mbps47.7 Mbps
bench-validator-10validator42.6 Mbps56.0 Mbps66.8 Mbps66.8 Mbps

Network received

instance/network/received_bytes_count rate; summaries over the mixed phase

Table view
NodeClassMeanp95Peak 60-s meanMax
bench-observer-11observer2.2 Mbps2.6 Mbps3.1 Mbps3.1 Mbps
bench-validator-01validator4.6 Mbps5.7 Mbps6.0 Mbps6.0 Mbps
bench-validator-02validator5.4 Mbps8.2 Mbps9.8 Mbps9.8 Mbps
bench-validator-03validator7.9 Mbps17.1 Mbps25.4 Mbps25.4 Mbps
bench-validator-04validator5.3 Mbps6.0 Mbps8.0 Mbps8.0 Mbps
bench-validator-05validator6.7 Mbps24.1 Mbps27.4 Mbps27.4 Mbps
bench-validator-06validator5.0 Mbps6.0 Mbps6.3 Mbps6.3 Mbps
bench-validator-07validator5.0 Mbps6.3 Mbps7.1 Mbps7.1 Mbps
bench-validator-08validator5.1 Mbps6.1 Mbps8.2 Mbps8.2 Mbps
bench-validator-09validator5.3 Mbps6.1 Mbps6.6 Mbps6.6 Mbps
bench-validator-10validator4.9 Mbps5.7 Mbps6.0 Mbps6.0 Mbps

Consensus health

MetricQueryClassNodesMedian of node meansWorst peak 60-s meanMax
tn_primary_commit_latency_secondsp95validator1010.8 s30.0 s30.0 s
tn_engine_execution_duration_secondsp95validator1016.2 s30.0 s30.0 s
tn_engine_execution_duration_secondsp95observer111.5 s26.0 s29.0 s
tn_worker_batch_size_bytesp95validator10243.9 KiB246.4 KiB246.4 KiB
tn_worker_batch_transactionsp95validator104941K1K
tn_executor_output_transactionsp95validator101K1K1K
tn_primary_consensus_output_bytesmaxvalidator101.0 MiB3.3 MiB5.2 MiB
tn_primary_consensus_output_bytesmaxobserver1958.1 KiB2.3 MiB5.2 MiB
tn_engine_queued_outputsmaxvalidator102.3188
tn_engine_queued_outputsmaxobserver11.35.758
tn_batch_builder_pending_pool_transactionsmaxvalidator105649.85K9.85K
tn_batch_builder_pending_pool_transactionsmaxobserver1000
reth_process_resident_memory_bytesmaxvalidator105.04 GiB6.69 GiB6.73 GiB
reth_process_resident_memory_bytesmaxobserver12.11 GiB3.66 GiB3.75 GiB
tn_primary_subdags_committed_totalratevalidator100.2050.4890.489
tn_primary_subdags_committed_totalrateobserver1000
tn_primary_round_laglagvalidator102.263234
tn_primary_round_laglagobserver1000

Commit latency p95

tn_primary_commit_latency_seconds

Round lag

tn_primary_round − tn_primary_committed_round

Engine queued outputs

tn_engine_queued_outputs, max by node

On-node sampler

NodeClassPeak RssAnonPeak RssFilePeak cachePSI memPSI ioPSI cpudb growthstatic_files growthRoot usedOOM Δ
bench-observer-11observer2.46 GiB1.50 GiB192 MiB 18.8 %0.9 %16.1 %1.8 %0 B1.70 GiB11.0 %0
bench-validator-01validator5.22 GiB1.59 GiB832 MiB 81.3 %2.4 %15.1 %14.5 %0 B1.70 GiB11.0 %0
bench-validator-02validator5.71 GiB1.59 GiB832 MiB 81.3 %3.0 %18.6 %16.7 %0 B1.70 GiB12.0 %0
bench-validator-03validator5.67 GiB1.65 GiB1.00 GiB 100.0 %3.1 %27.3 %11.6 %0 B1.70 GiB11.0 %0
bench-validator-04validator5.93 GiB1.61 GiB832 MiB 81.3 %3.5 %29.6 %12.0 %0 B1.70 GiB11.0 %0
bench-validator-05validator5.43 GiB1.57 GiB896 MiB 87.5 %0.9 %36.1 %12.0 %0 B1.61 GiB11.0 %0
bench-validator-06validator5.46 GiB1.58 GiB896 MiB 87.5 %3.4 %14.8 %15.0 %0 B1.70 GiB11.0 %0
bench-validator-07validator5.80 GiB1.69 GiB832 MiB 81.3 %3.3 %18.1 %17.7 %0 B1.70 GiB12.0 %0
bench-validator-08validator5.40 GiB1.69 GiB832 MiB 81.3 %1.7 %17.8 %14.5 %0 B1.70 GiB11.0 %0
bench-validator-09validator5.37 GiB1.68 GiB832 MiB 81.3 %1.3 %33.9 %16.1 %0 B1.70 GiB11.0 %0
bench-validator-10validator5.57 GiB1.60 GiB832 MiB 81.3 %4.0 %33.1 %13.2 %0 B1.70 GiB11.0 %0

RssAnon vs RssFile

/proc/<pid>/status of the telcoin process on the node with the largest RSS

consensus-db/cache size

du -sb of consensus-db/cache vs the 1 GiB ceiling

Pressure stall (some, avg10)

/proc/pressure on the node with the highest memory PSI

Headroom

Per role and per node, from sizing.json (p95 over the landed window; PSI and RssAnon max over the whole sampler run). The lines follow the hardware doc's tier rules: CPU p95 over 70 % of vCPUs; memory used p95 over 75 % of RAM, or telcoin RSS p95 over 85 % of RAM; disk write p95 over 60 % of the PD MiB/s limit, or write IOPS p95 over 60 % of the IOPS limit; IO pressure over 20 % or memory pressure over 5 % (PSI some avg10, max); network p95 over 50 % of 1 Gbps; root disk over 80 %. A line crossed on any node puts the role at the limit; a nonzero reading within 15 points of its line is tight; otherwise the role has headroom. OOM kills, disk throttling and degraded disk performance are flagged too.

SignalClassWorst nodeValueShare of capacityLineNodes over / tightStatus
CPU p95validatorbench-validator-0355.2 %55.2 %> 70 %0 / 1 tight
Memory used p95validatorbench-validator-0780.8 %80.8 %> 75 %6 / 4 at the limit
telcoin RSS p95validatorbench-validator-047.15 GB83.3 %> 85 %0 / 10 tight
PSI memory maxvalidatorbench-validator-104.0 %4.0 %> 5 %0 / 10 tight
RssAnon peakvalidatorbench-validator-046.36 GB74.1 %info– info
Disk write p95validatorbench-validator-09137 MB/s54.4 %> 60 %0 / 4 tight
Write IOPS p95validatorbench-validator-045,11956.9 %> 60 %0 / 8 tight
PSI io maxvalidatorbench-validator-0937.2 %37.2 %> 20 %5 / 5 at the limit
Root disk maxvalidatorbench-validator-0611.5 %11.5 %> 80 %0 / 0 headroom
Network p95validatorbench-validator-1056 Mbps5.6 %> 50 %0 / 0 headroom
Consensus cache peakvalidatorbench-validator-031.00 GiB100.0 %info– info
OOM killsvalidator––anyany– none
Disk throttlingvalidator––metric type unavailable in projectany– not collected
Disk performance degradedvalidator––any sampleany– none
CPU p95observerbench-observer-1127.4 %27.4 %> 70 %0 / 0 headroom
Memory used p95observerbench-observer-1137.9 %37.9 %> 75 %0 / 0 headroom
telcoin RSS p95observerbench-observer-113.71 GB43.2 %> 85 %0 / 0 headroom
PSI memory maxobserverbench-observer-111.0 %1.0 %> 5 %0 / 1 tight
RssAnon peakobserverbench-observer-112.64 GB30.7 %info– info
Disk write p95observerbench-observer-1163 MB/s25.2 %> 60 %0 / 0 headroom
Write IOPS p95observerbench-observer-114,76552.9 %> 60 %0 / 1 tight
PSI io maxobserverbench-observer-1116.1 %16.1 %> 20 %0 / 1 tight
Root disk maxobserverbench-observer-1110.6 %10.6 %> 80 %0 / 0 headroom
Network p95observerbench-observer-113 Mbps0.3 %> 50 %0 / 0 headroom
Consensus cache peakobserverbench-observer-110.19 GiB18.8 %info– info
OOM killsobserver––anyany– none
Disk throttlingobserver––metric type unavailable in projectany– not collected
Disk performance degradedobserver––any sampleany– none

Per node

NodeClassCPU p95Memory used p95telcoin RSS p95PSI memory maxRssAnon peakDisk write p95Write IOPS p95PSI io maxRoot disk maxNetwork p95Consensus cache peak
Line> 70 %> 75 %> 85 %> 5 %info> 60 %> 60 %> 20 %> 80 %> 50 %info
bench-validator-01validator48.8 %73.8 %6.98 GB 81 %2.4 %5.63 GB 66 %75 MB/s 30 %4,838 54 %15.1 %11.4 %43 Mbps 4 %0.94 GiB 94 %
bench-validator-02validator48.6 %79.8 %7.09 GB 82 %3.0 %6.13 GB 71 %79 MB/s 31 %4,676 52 %19.9 %11.4 %51 Mbps 5 %0.94 GiB 94 %
bench-validator-03validator55.2 %78.6 %7.13 GB 83 %3.1 %6.08 GB 71 %107 MB/s 43 %3,248 36 %27.3 %11.3 %46 Mbps 5 %1.00 GiB 100 %
bench-validator-04validator55.0 %80.3 %7.15 GB 83 %3.5 %6.36 GB 74 %120 MB/s 48 %5,119 57 %29.6 %11.5 %45 Mbps 4 %1.00 GiB 100 %
bench-validator-05validator45.8 %74.2 %6.85 GB 80 %0.9 %5.83 GB 68 %128 MB/s 51 %3,554 39 %36.1 %11.2 %54 Mbps 5 %1.00 GiB 100 %
bench-validator-06validator49.6 %76.4 %7.03 GB 82 %3.4 %5.86 GB 68 %75 MB/s 30 %4,712 52 %15.4 %11.5 %56 Mbps 6 %1.00 GiB 100 %
bench-validator-07validator52.5 %80.8 %7.08 GB 82 %3.3 %6.23 GB 72 %78 MB/s 31 %4,789 53 %18.1 %11.3 %46 Mbps 5 %0.94 GiB 94 %
bench-validator-08validator52.9 %74.3 %6.98 GB 81 %1.7 %5.79 GB 67 %77 MB/s 31 %4,674 52 %17.8 %11.3 %49 Mbps 5 %0.94 GiB 94 %
bench-validator-09validator47.1 %74.1 %7.09 GB 83 %1.3 %5.77 GB 67 %137 MB/s 54 %4,192 47 %37.2 %11.4 %45 Mbps 4 %0.94 GiB 94 %
bench-validator-10validator48.8 %77.8 %7.15 GB 83 %4.0 %5.98 GB 70 %129 MB/s 51 %4,551 51 %33.1 %11.5 %56 Mbps 6 %0.94 GiB 94 %
bench-observer-11observer27.4 %37.9 %3.71 GB 43 %1.0 %2.64 GB 31 %63 MB/s 25 %4,765 53 %16.1 %10.6 %3 Mbps 0 %0.19 GiB 19 %

Small figures are the share of capacity: RAM for RSS, the PD limit for disk write (VM cap applied), 1 Gbps for network. Hover a cell for its JSON path.

Verdict

Fleet A vs Fleet B

A = Fleet A · e2-custom-4-8192 · r20260923-0708
B = Fleet B · c3-highcpu-8 · r20260923-0515

Verdict

at the limit e2-custom-4-8192 validators at 3,569 sustained TPS: at the limit on memory and disk. Memory at the limit: memory used p95 76.4–80.8 % of RAM, over the 75 % line on 6 of 10 validators (bench-validator-02, 03, 04, 06, 07, 10); telcoin RSS p95 7.15 GB, 83.3 % of 8 GiB RAM (bench-validator-04), tight against the 85 % line; memory pressure (PSI some avg10) max 4.0 % (bench-validator-10), tight against the 5 % line. Disk at the limit: IO pressure (PSI some avg10) max 27.3–37.2 %, over the 20 % line on 5 of 10 validators (bench-validator-03, 04, 05, 09, 10); write p95 137 MB/s, 54.4 % of the 240 MiB/s limit (bench-validator-09), tight against the 60 % line; write IOPS p95 5,119, 56.9 % of the 9,000 IOPS limit (bench-validator-04), tight against the 60 % line. CPU tight: CPU p95 55.2 % of 4 vCPUs (bench-validator-03), tight against the 70 % line. Network has headroom: network p95 56 Mbps, 5.6 % of 1 Gbps (bench-validator-10). Next step: add RAM; a faster volume (pd-ssd adds 30 IOPS and 0.48 MiB/s per GB up to the per-VM cap; past the cap, a larger VM).

tight e2-custom-4-8192 observers at 3,569 sustained TPS: tight on memory and disk. Memory tight: memory pressure (PSI some avg10) max 1.0 % (bench-observer-11), tight against the 5 % line. Disk tight: write IOPS p95 4,765, 52.9 % of the 9,000 IOPS limit (bench-observer-11), tight against the 60 % line; IO pressure (PSI some avg10) max 16.1 % (bench-observer-11), tight against the 20 % line. CPU has headroom: CPU p95 27.4 % of 4 vCPUs (bench-observer-11). Network has headroom: network p95 3 Mbps, 0.3 % of 1 Gbps (bench-observer-11).

at the limit c3-highcpu-8 validators at 5,384 sustained TPS: at the limit on disk. Disk at the limit: write p95 181–190 MB/s, up to 75.5 % of the 240 MiB/s limit, over the 60 % line on all 10 validators (bench-validator-01, 02, 03, 04, 05, 06, 07, 08, 09, 10); IO pressure (PSI some avg10) max 46.7–59.8 %, over the 20 % line on all 10 validators (bench-validator-01, 02, 03, 04, 05, 06, 07, 08, 09, 10). CPU has headroom: CPU p95 24.6 % of 8 vCPUs (bench-validator-04). Memory has headroom: memory used p95 52.0 % of RAM (bench-validator-04); telcoin RSS p95 10.78 GB, 62.8 % of 16 GiB RAM (bench-validator-04); memory pressure (PSI some avg10) max 0.0 % (bench-validator-01). Network has headroom: network p95 86 Mbps, 8.6 % of 1 Gbps (bench-validator-05). Next step: a faster volume (pd-ssd adds 30 IOPS and 0.48 MiB/s per GB up to the per-VM cap; past the cap, a larger VM).

at the limit c3-highcpu-8 observers at 5,384 sustained TPS: at the limit on disk. Disk at the limit: write p95 174 MB/s, 69.2 % of the 240 MiB/s limit on bench-observer-11, over the 60 % line; IO pressure (PSI some avg10) max 56.6 % on bench-observer-11, over the 20 % line. CPU has headroom: CPU p95 13.1 % of 8 vCPUs (bench-observer-11). Memory has headroom: memory used p95 15.4 % of RAM (bench-observer-11); telcoin RSS p95 3.65 GB, 21.2 % of 16 GiB RAM (bench-observer-11); memory pressure (PSI some avg10) max 0.0 % (bench-observer-11). Network has headroom: network p95 4 Mbps, 0.4 % of 1 Gbps (bench-observer-11). Next step: a faster volume (pd-ssd adds 30 IOPS and 0.48 MiB/s per GB up to the per-VM cap; past the cap, a larger VM).

Validators: e2-custom-4-8192 is at the limit on memory and disk, tight on CPU, with headroom on network; c3-highcpu-8 is at the limit on disk, with headroom on CPU, memory and network. The same disk limit binds both fleets: every validator runs a 100 GB pd-ssd, which allows 9,000 IOPS on both; its throughput limit is 240 MiB/s on both. Disk is at the limit on both (e2-custom-4-8192: IO pressure (PSI some avg10) max 37.2 %; c3-highcpu-8: write p95 190 MB/s, 75.5 % of the 240 MiB/s limit, IO pressure (PSI some avg10) max 59.8 %). Changing the machine shape did not clear it; the volume is the binding resource. c3-highcpu-8 stays under the memory lines that e2-custom-4-8192 crosses.

Observers: e2-custom-4-8192 is tight on memory and disk, with headroom on CPU and network; c3-highcpu-8 is at the limit on disk, with headroom on CPU, memory and network. Disk is at the limit on c3-highcpu-8 only. e2-custom-4-8192 stays under the disk lines that c3-highcpu-8 crosses.

KPIFleet AFleet BB ÷ A
Sustained mixed TPS (own)3,569.4 tx/s5,383.5 tx/s1.51×
Sustained mixed TPS (chain)3,569.4 tx/s5,383.5 tx/s1.51×
Landed / accepted99.6 %99.5 %1.00×
Reverted (own)00–
Latency p507,595 ms7,677 ms1.01×
Latency p9542,041 ms31,481 ms0.75×
Gas / s (own)113M gas/s118M gas/s1.05×
Capacity at mix4,268.6 tx/s4,406.0 tx/s1.03×

Headroom, worst node per class

SignalClassFleet AFleet BLine
CPU p95validator55.2 % tight24.6 % headroom> 70 %
Memory used p95validator80.8 % at the limit52.0 % headroom> 75 %
telcoin RSS p95validator7.15 GB 83.3 % tight10.78 GB 62.8 % headroom> 85 %
PSI memory maxvalidator4.0 % tight0.0 % headroom> 5 %
Disk write p95validator137 MB/s 54.4 % tight190 MB/s 75.5 % at the limit> 60 %
Write IOPS p95validator5,119 56.9 % tight2,440 27.1 % headroom> 60 %
PSI io maxvalidator37.2 % at the limit59.8 % at the limit> 20 %
Root disk maxvalidator11.5 % headroom12.2 % headroom> 80 %
Network p95validator56 Mbps 5.6 % headroom86 Mbps 8.6 % headroom> 50 %
CPU p95observer27.4 % headroom13.1 % headroom> 70 %
Memory used p95observer37.9 % headroom15.4 % headroom> 75 %
telcoin RSS p95observer3.71 GB 43.2 % headroom3.65 GB 21.2 % headroom> 85 %
PSI memory maxobserver1.0 % tight0.0 % headroom> 5 %
Disk write p95observer63 MB/s 25.2 % headroom174 MB/s 69.2 % at the limit> 60 %
Write IOPS p95observer4,765 52.9 % tight1,788 19.9 % headroom> 60 %
PSI io maxobserver16.1 % tight56.6 % at the limit> 20 %
Root disk maxobserver10.6 % headroom11.1 % headroom> 80 %
Network p95observer3 Mbps 0.3 % headroom4 Mbps 0.4 % headroom> 50 %

Own TPS, trailing 60-s mean

aligned on each fleet's mixed-phase start

CPU, hottest validator

aligned on each fleet's mixed-phase start

Memory used, hottest validator

aligned on each fleet's mixed-phase start

telcoin RSS, hottest validator

aligned on each fleet's mixed-phase start

Disk MiB/s, hottest validator

aligned on each fleet's mixed-phase start

Network sent, hottest validator

aligned on each fleet's mixed-phase start

Reference chains

Not collected: reference chains.