rustbgpd
- Withdraw p50
- 0.28–0.48 s
- Re-announce p50
- 0.46–0.49 s
- withdraw p50
- 0.00s
- re-announce
- 0.00s
rustbgpd
A BGP route server and route reflector in Rust.
cargo install rustbgpd
Animated topology of a BGP flapstorm. A steel withdrawal wave, then a Ferris-orange re-announce flood. Withdraw p50 0.48 seconds. Re-announce 0.49 seconds. Press R or use Replay to run it again.
01
Flapstorm propagation
700 clients × 400,400 routes, same host, same harness. 50 members hard-close simultaneously; 650 survivors observe 28,600 withdrawn prefixes. The 50 reconnect and re-announce.
| Daemon | Withdraw p50 | Re-announce p50 |
|---|---|---|
| rustbgpd | 0.28–0.48 s | 0.46–0.49 s |
| BIRD 3.3.1 | 0.45–0.61 s | 2.84–4.22 s |
| OpenBGPD 9.1 | 10.8–12.4 s | 21.2–21.7 s |
BIRD 3.3.1
- Withdraw p50
- 0.45–0.61 s
- Re-announce p50
- 2.84–4.22 s
OpenBGPD 9.1
- Withdraw p50
- 10.8–12.4 s
- Re-announce p50
- 21.2–21.7 s
The re-announce clock starts only after all 50 sessions re-reach Established, so it measures re-delivery fan-out (~18.6M NLRI), not reconnect pacing. [rustbgpd source receipt] · [BIRD 3.3.1 source receipt] · [OpenBGPD 9.1 source receipt]
01b — SIGHUP policy swap
Reload performance
700 clients × 400,400 routes, same host, same harness. SIGHUP policy reload stall and completion times.
| Daemon | Stall p50 | Stall worst | Completion p50 | Completion worst |
|---|---|---|---|---|
| rustbgpd | 0.43–0.84 s | 1.85 s | 1.5–2.2 s | 2.9 s |
| BIRD 3.3.1 | 1.61–2.29 s | 7.73 s | 77–86 s | 93.8 s |
| OpenBGPD 9.1 | 0.25–0.29 s | 0.36 s | 249–253 s | 252.9 s |
rustbgpd
- Stall p50
- 0.43–0.84 s
- Stall worst
- 1.85 s
- Completion p50
- 1.5–2.2 s
- Completion worst
- 2.9 s
BIRD 3.3.1
- Stall p50
- 1.61–2.29 s
- Stall worst
- 7.73 s
- Completion p50
- 77–86 s
- Completion worst
- 93.8 s
OpenBGPD 9.1
- Stall p50
- 0.25–0.29 s
- Stall worst
- 0.36 s
- Completion p50
- 249–253 s
- Completion worst
- 252.9 s
Stall = time during which no prefix is reachable (withdrawn but not yet re-announced). Completion = full table restored. [rustbgpd source receipt] · [BIRD 3.3.1 source receipt] · [OpenBGPD 9.1 source receipt]
02
24-hour soak receipts
Both green, both archived in-repo. No control daemon ran alongside them.
Route-server shape
PASS1,000 eBGP clients × 400,000 routes, 24 h
- SIGHUP policy reloads
- 48
- Max-prefix trip → timed-restart cycles
- 6
- Late-window RSS slope
- 0.0724 MB/h
Route-reflector shape
PASS1,000 iBGP clients × 100,000 routes, 24 h
- Churn cycles
- 5,486,092
- Session flaps
- 0
- Terminal reflected prefixes
- 99,900
- Terminal parse errors
- 0
03
Honesty
The receipts also include the losses.
OpenBGPD winsReload stall
OpenBGPD's stall p50 is 0.25–0.29 s vs rustbgpd's 0.43–0.84 s at 700 × 400,400. [source receipt]
BIRD winsMemory use
BIRD wins settled memory outright (~400 MiB vs ~760 MiB). S2 settled memory against BIRD is a dead heat; BIRD wins S3 settled outright. [source receipt]
What this is and is not
- Public alpha at v0.64.0. Not production-grade yet.
- Linux only. No FreeBSD, no Mac, no Windows.
- EVPN is functional but outside the v1 compatibility promise.
- Every receipt is currently self-generated — no external pilot yet.
- OpenBGPD has a smaller reload stall; BIRD uses less memory; S2 settled memory against BIRD is a dead heat; BIRD wins S3 settled outright.