Reliable buffer too small for the round-trip
Rule 32 · applies to the publisher · Back to all rules
Breaks a guarantee. The send buffer fills before acknowledgments return, so the writer cannot retransmit and reliable delivery drops samples.
- Settings involved: Resource Limits and Reliability
- What QoS Guard checks:
[RELIABLE] ∧ [HIST.kind=KEEP_ALL] ∧ [RESLIM.mpi < ⌈2×RTT/PP⌉+1]
Example
At 100 ms RTT and a 20 ms period you need about 11 slots, but max_samples_per_instance is 4. Delivery collapses under loss.
How to fix it
Raise max_samples_per_instance to at least ceil(2 x RTT / PP) + 1 for reliable KEEP_ALL over lossy links.
Why this rule is flagged
What the DDS specification says
The DDS specification does not settle this case on its own, so the rule rests on direct measurement.
What the engine source code shows
The behavior here does not depend on a specific engine's implementation, so the rule follows from the measurements.
What the measurements show
| Item | Value |
|---|---|
| Dataset | Download CSV |
| Fixed QoS setting | RELIAB = RELIABLE, overwrite-mode drop_kl |
| Tested variable | depth_bound |
| Tested values | depth_bound ∈ {1, 2, 3, 4, 5, 6, 7, 8, 10, 11, 12, 14, 18} |
| Rule boundary | ceil(2 × RTT / PP) + 1 |
| Boundary values | RTT = 50 ms, PP = 20 ms → threshold = 6; RTT = 100 ms, PP = 20 ms → threshold = 11 |
| Tested engine / version | Cyclone DDS 0.10.5 |
| Network setting | RTT ∈ {50 ms, 100 ms}, loss ∈ {10%, 20%}, PP = 20 ms, message size = 1024 B |
Measurement result
The overwrite-mode measurement shows Cyclone DDS drop_kl delivery behavior across tested depth_bound values, with the rule boundary computed as ceil(2 × RTT / PP) + 1.