Fleet Performance

Pathfinder · the client fleet exploring self-avoiding random walks
compiled 2026-09-13 16:52
2d ago · refresh 300s
1 client could not be reached at build time — their performance.log was unreadable. They are flagged, not dropped, in §V.

4 of 5 client machines are reporting, running 38 worker processes between them. Together they have explored 508.39B walks and kept 9.2K as interesting — about 1 in 55.3M — over 33,834.0 compute-hours. The fleet is presently turning at 908.73M paths an hour. What follows is its portrait: the work accumulated over time, how each machine ranks, how efficiently each worker runs, and whether adding more workers would still buy more throughput.

Paths explored
508.39B
508,393,443,508 across the fleet
Interesting
9.2K
0.0000% hit rate
Throughput
908.73M /hr
peak 38.88M/hr per worker
Clients
4 / 5
1 unreachable
Compute time
1409d 18h
summed worker runtime
Paths collected
54.1K
saved on the server
I

The work over time

cumulative paths since the first checkpoint

Paths explored across the whole fleet, accumulated from the earliest logged checkpoint to the latest.

0200B400B600B800B1TJul 5, 06:01Aug 8, 17:27Sep 13, 16:51
Fig. 1. Cumulative paths against wall-clock time. The blue curve is the fleet total; switch to By client for one ink curve per machine, each labelled at its head.
II

Paths explored by client

total output per machine
040B80B120B160B200BPathfinder BPathfinder B: 170,721,432,996 paths170.72BPathfinder CPathfinder C: 169,007,172,185 paths169.01BPathfinder APathfinder A: 122,548,838,327 paths122.55BA New BeginningA New Beginning: 46,116,000,000 paths46.12B
III

Worker efficiency

throughput per worker — the fair race
Fleet throughput
908.73M /hr
Per worker (avg)
23.91M /hr
Projected / day
21.81B
Median time / path
237.4 µs
010M20M30M40M50MA New BeginningA New Beginning: 37,297,086 paths/hr per worker37.3MPathfinder BPathfinder B: 17,556,503 paths/hr per worker17.56MPathfinder CPathfinder C: 13,298,708 paths/hr per worker13.3MPathfinder APathfinder A: 12,531,435 paths/hr per worker12.53M
Table 1. Per-client efficiency, derived from performance.log alone. Stability is 1 − the throughput coefficient of variation.
ClientWorkersThroughput /hrPer worker /hr Time / pathHit rateStabilityProj. / day
A New Beginning 16 596,753,382 37,297,086 94.7 µs 0.0000% 99% 14.32B
Pathfinder B 6 105,339,020 17,556,503 206.2 µs 0.0000% 95% 2.53B
Pathfinder C 8 106,389,661 13,298,708 268.6 µs 0.0000% 89% 2.55B
Pathfinder A 8 100,251,478 12,531,435 287.7 µs 0.0000% 99% 2.41B
IV

Scaling

would more workers add throughput?
010M20M30M40M50M02.557.51012.51517.520Worker processes per clientThroughput per worker (paths/hr)Pathfinder A: 8 workers, 12.53M/hr per workerPathfinder APathfinder B: 6 workers, 17.56M/hr per workerPathfinder BPathfinder C: 8 workers, 13.3M/hr per workerPathfinder CA New Beginning: 16 workers, 37.3M/hr per workerA New Beginning

Near-linear scaling — headroom to add workers

Per-worker throughput holds roughly steady as worker count rises, so each added worker contributes close to a full worker's output. Adding more workers (or clients) should increase total throughput proportionally.
Signals: per-worker vs workers r = 0.92, total vs workers r = 0.98, avg 20.17M/hr per worker.

V

The fleet, machine by machine

every client, reachable or not
Table 2. Each client's performance.log as read at build time. The sparkline traces throughput (paths/hr) across the log; unreachable clients are flagged, not dropped.
ThroughputClientStatusWorkers PathsInterestingRateNow /hrPeak /hr ComputeLast entry
Pathfinder A 10.0.0.33 active 8 122.55B 2.2K 0.000% 100.25M 14.02M 409d 2h 2d ago
Pathfinder B 10.0.0.31 active 6 170.72B 3.1K 0.000% 105.34M 25.84M 402d 12h 2d ago
Pathfinder C 10.0.0.10 active 8 169.01B 3.1K 0.000% 106.39M 26.66M 547d 10h 2d ago
— HP Laptop 10.0.0.32 unreachable — — — — — — — no suitable wrapper could be found
A New Beginning 10.0.0.3 stale 16 46.12B 816 0.000% 596.75M 38.88M 50d 18h 60d ago
Fleet total4/5 38 508.39B 9.2K 0.0000% 908.73M 38.88M 1409d 18h 2026-09-13 16:51
Compiled from the fleet telemetry · schema v1 · reads each client's performance.log. Set in Palatino. All measurement, no decoration.