Cluster FLEET-07 6 NODES · 22 GPU
6 ONLINE
2 DEGRADED
7 ALERTS
6.28 kW
--:--:--
◉ LIVE FLEET THROUGHPUT · all decode slots
0tok/s
▲ 0.0% · 60s
SLA ≥ 600 tok/s
decode slots14
prefill queue2.4k tok
spec. accept rate0.83
median ttft118 ms
Cluster mesh
10.0.4.0/22 · RoCE
online
degraded
maintenance
link 25 Gb/s avg
POWER DRAW · PUE 1.14
0.00kW
GPU bundle0.00 kW
CPU / mobo0.00 kW
cooling loop0.00 kW
cost / hour$0.74
circ temp21.4 °C
QUEUE PRESSURE
active0
waiting0
done 1h0
failed0
drain rate—
Telemetry interactive · hover for values
Job completion time
p50 / p95 per model · last 15 minp50p95
Active alerts
7 openServer nodes
Fleet inventory searchable · sortable · servers + models
| Name ▲ | Type | Status | Node | Quant / Cores | Ctx / Disk | tok/s | Jobs | Uptime / VRAM |
|---|
GPU monitor
Utilization heatmap
30 min · per nodeutil %temp
Model deployments
| Model | Quant | Context | GPU assignment | Status | Jobs | VRAM | tok / s | p95 |
|---|