Node benchmark board

Verified YABS runs. Averages roll on a 6-month cutoff (run date), so stale numbers drop out.

Another node: curl -fsSLO https://05877bc1b477317f.mcx.fi/yabs/probe.sh && sh probe.sh registers it and schedules a monthly run. sh probe.sh daily|weekly|monthly changes that schedule. Then sh probe.sh now runs one now, sh probe.sh status shows progress.

From a node with the admin token: yabs.sh -j -s https://05877bc1b477317f.mcx.fi/yabs/?token=TOKEN&node=$(hostname -s)

README · llms.txt (machine-readable install and API summary) · JSON · 3D hangar · history · source

Add a run (paste YABS JSON)

Node name(s), comma-separated:   Upload token:

Nodes

nodecpucoresramswapdiskgpuosvmlast run
le4-0-66landstalkerAMD Ryzen 5 3500X 6-Core Processor6125.8 GiB29.8 GiB93.6 TB—Debian GNU/Linux 12 (bookworm)NONE2026-10-27
le4-0-78-102carolineAMD EPYC 7451 24-Core Processor1621.1 GiB1.0 GiB27.3 TB—Debian GNU/Linux 12 (bookworm)KVM2026-10-27
le6-3-58ferosIntel(R) N100431.1 GiB54.7 GiB126.9 TBrenderD128Debian GNU/Linux 12 (bookworm)NONE2026-10-27

Averages, last 6 months

noderuns4k MB/s4k IOPS64k512k1msend Gbit/srecv Gbit/sGB s/m
le4-0-66landstalker12.72869826.87173.5204.56.457.08—
le4-0-78-102caroline20.353903.37024.0534.037.6413.05—
le6-3-58feros10.227576.85835.1542.991.962.30—

Per-node runs (last 6 months)

le4-0-66landstalker

date4k MB/sIOPS64k512k1msend Gbit/srecv Gbit/syabs
2026-10-272.72869826.87173.5204.56.457.08v2026-10-07

le4-0-78-102caroline

date4k MB/sIOPS64k512k1msend Gbit/srecv Gbit/syabs
2026-10-270.383975.39534.5448.387.5812.60v2026-10-07
2026-10-270.322821.34513.5519.677.7013.50v2026-10-07

le6-3-58feros

date4k MB/sIOPS64k512k1msend Gbit/srecv Gbit/syabs
2026-10-270.227576.85835.1542.991.962.30v2026-10-07

Disk numbers are the fio mixed read/write figures (speed_rw, IOPS). Net numbers are the best iperf3 location, since a timed-out location still reports a few Kbit/s. On shared storage these swing widely between runs: read them as a trend, not a spec.