Header Overhead and the Cost of a Missing Backup Route
Ada re-derives this chapter’s own numbers step by step, at full precision
ADA · CALCULATION AUDIT
Header Overhead and the Cost of a Missing Backup Route
An 8-hop DSR route with 2-byte node IDs wraps a 20-byte sensor reading in 16 bytes of route list — roughly 44 percent of the packet — and when that route goes stale the repair costs 315 ms, falling to 115 ms if a backup is already cached. Both costs look like overhead worth trimming. This audit re-derives the header share and the repair timeline and asks whether a long-path, tight-deadline deployment should spend its budget shaving route bytes or maintaining that one backup route.
Companion to the chapter DSR Worked Examples — every number here comes from that chapter.
Ada: This section makes two claims I can pin down with the chapter’s own numbers — that an 8-hop source route is “about 44 percent” of a small packet, and that a stale route costs “roughly 315 ms” to repair. Let me verify both, then push each one step further.
The header first. An 8-hop route with 2-byte node IDs and a 20-byte payload:
- Route list:
8 x 2 = 16 bytes - Packet, ignoring other headers:
16 + 20 = 36 bytes - Header share:
16 / 36 = 0.44444... = 44.4%
So “about 44 percent” is right. That share is not a fixed property of DSR — it is set by the ratio of route bytes to payload. For the header to fall below a quarter of the packet, the payload must grow past 3 x 16 = 48 bytes, because 16 / (16 + 48) = 0.25. That is the whole reason DSR punishes small telemetry: a 20-byte reading cannot dilute a 16-byte route, but a 48-byte one nearly can.
Now the repair. The chapter sums detection, error propagation, and rediscovery:
- With a fresh discovery:
100 + 15 + 200 = 315 ms - With a valid alternate already cached:
100 + 15 = 115 ms(the 200 ms discovery is skipped)
The cached alternate removes 315 - 115 = 200 ms, or 200 / 315 = 63.5% of the repair time. That one backup route is worth far more than any header optimization here: it cuts nearly two-thirds off the delay spike, while trimming the header saves at most a few bytes.
The design lesson is that DSR’s two costs pull in opposite directions — a longer path inflates every packet’s header, but the expensive event is the missing backup route, so a deployment with long paths and strict deadlines should spend its budget on maintaining alternates, not on shaving route bytes.
Every number above is taken from the chapter’s own material and re-derived step by step.