The RAM Buffer Covers Only a Quarter of the Outage
Ada re-derives this chapter’s own numbers step by step, at full precision
ADA · CALCULATION AUDIT
The RAM Buffer Covers Only a Quarter of the Outage
A phone or vehicle gateway loses its cellular uplink in tunnels and dead zones, so it must buffer field data until the link returns. The chapter’s worked example has sensors delivering 20 messages per second at 200 bytes each — 4 KB/s — against a worst outage of 15 minutes (900 s), which needs a 3.6 MB buffer to lose nothing. But a 1 MB RAM queue overflows in just 250 s, about four minutes. This audit re-runs that sizing to ask how much of the outage a plain RAM buffer really survives.
Companion to the chapter Lab: Mobile Gateway Protocols — every number here comes from that chapter.
Ada: The practitioner section sizes a buffer for a tunnel. I want to confirm the arithmetic and then measure how much of the worst outage a plain RAM queue actually survives, because that gap is what forces the flash-or-drop decision.
Start from the field ingest rate and the 15-minute (900 s) worst outage:
- Ingest:
20 msg/s x 200 bytes = 4000 bytes/s = 4 KB/s - Buffer to cover the whole gap:
4 KB/s x 900 s = 3600 KB = 3.6 MB - A 1 MB RAM queue lasts:
1000 KB / 4 KB/s = 250 s, or250 / 60 = 4.2 minutes
So the RAM-only design holds 250 / 900 = 27.8% of the worst outage before it overflows, barely a quarter. The remaining 3600 - 1000 = 2600 KB has to land in flash, or a drop policy has to discard it on purpose and count what it threw away. That 28 percent figure is the whole point of the lab: a buffer that looks fine on a bench with steady connectivity silently loses nearly three-quarters of a real tunnel’s data unless the outage duration, not the average, is what sizes it.
Every number above is taken from the chapter’s own material and re-derived step by step.