Commander testing
Commander
testing.
Wrapped.
22 – 24 April 2026 · servicom deployment
84,377 commands
99.99% success
p50 latency 464 ms
Zero downtime
Three days. One fleet of MOTOTRBO radios. A lot of CSBKs. Scroll on — we'll show you exactly how Commander performed on your watch.
In just 57 hours of testing
You pushed 84,377
radio commands
through Commander.
That's one command every 2.4 seconds, sustained. Stun, unstun, repeat. The kind of soak test that turns lesser systems into cautionary tales.
Total commands completed
84,377
42,187 stuns · 42,190 unstuns · all acknowledged by the radio over the air.
Commands per hour
22 Apr 08:0024 Apr 15:00 UTC
Success rate
99.99%
Out of 84,377 commands over three days, only 8 were outright declined by the radio. Every other command was either acknowledged first time or correctly queued for later by Commander. That's the kind of reliability you design for, not hope for.
Your command mix
Perfectly balanced.
You hit stun and unstun almost to the command. Exactly the kind of symmetric load a production muster would throw at Commander on a bad morning.
Your peak hour
1,730
commands
in an hour.
At 20:00 UTC on 23 April 2026, Commander ran for a full 60 minutes at one command every two seconds. Every single one completed end-to-end. No queue backlog, no dropped frames, no drama.
The fleet you exercised
23
radios.
Twenty-three production DMR radios, each hammered for three days straight. Commander's queue kept the load evenly distributed — every radio saw thousands of stun/unstun cycles without hogging the air.
Your most-called radio
#101
3,934 commands. Radio 101 took more stun/unstun pairs than any other in the fleet — and came back for more every time.
- 01Radio 1013,934
- 02Radio 1243,926
- 03Radio 2013,926
- 04Radio 2053,918
- 05Radio 2043,836
The latency waterfall
Three stages.
One blink.
Every command flows through the same three stages. Here's how the median 463 ms breaks down — and how tightly the tails hold at p95 and p99.
463 ms · p50 cumulative total
01 · Gateway
API → Redis stream → radio-gateway → ready to transmit.
02 · Over the air
CSBK fully transmitted on the MOTOTRBO wireline protocol.
03 · Radio ACK
Radio replied over-the-air, decoded, persisted, SSE fired.
Latency distribution (end-to-end)
99% under
700 milliseconds.
0 ms250 ms500 ms750 ms1 second
From the minimum at 278 ms through to the 99th percentile at 676 ms, every marker stays packed well inside the sub-second band. A gap of just 212 ms between p50 and p99 — that's tail behaviour you can build an SLA around.
Availability
No restarts. No blips.
No drama.
All systems nominal · 72 h observation window
22 Apr 00:0023 Apr 00:0024 Apr 00:0024 Apr 15:00 UTC
Five core services, minute by minute, across the whole test window. Every strip unbroken — Commander just kept running.
The repeater stack
The whole stack.
Held open.
Eight MOTOTRBO peers were exercised across a 3-day test run. The gateway held every peer alive throughout — and the architecture scales well beyond what this run put through it.
8
Peers exercised in this test run
474,930
Packets relayed through the gateway at peak
188,044
Keep-alive frames exchanged with the peer stack
57 h
Of continuous peer keep-alive observed
Thanks for testing, Manj.
Commander
delivered.
Every time.
84,377 commands. 8 declines. Zero downtime. 99.99% first-time success. Commander isn't prototype software — it's production-grade infrastructure for fleet-scale radio control.
Every figure on this page was pulled directly from live Grafana Cloud Loki logs — the servicom production deployment, 22 to 24 April 2026. No synthetic data, no rounding up.