Design a Bus Tracking & Route Optimization System - System Design | WinJob

Problem Statement: Public Bus Tracking Platform

Design city-scale bus tracking and route optimization: GPS ingest every 5–15s, map-matched progress along GTFS shapes, fused ETAs at stops, operator reroutes, and planner analytics. This section covers problem statement: public bus tracking platform in the understanding phase.

Operational detail

Quantify before naming SKUs: assume 8,000 buses, 5M daily riders, 10s GPS interval → ~800 ingest events/s average, 3× peak near commute. Store 7-day hot telemetry in Timescale, 90-day cold in object storage. CAP split: AP for live vehicle state (brief staleness OK), CP for schedule versions and alert publications.

Failure and edge cases

GPS dropout in urban canyons → fall back to schedule with STALE badge
Ghost vehicle after depot return → trip_id mismatch detector
Detour without GTFS-RT Alert → riders see impossible ETAs
Kafka lag spike → widen ETA window; never show negative countdown

Interview checkpoints

javaOne Dark Pro

1 public final class GtfsTripKey { public final String tripId; public final String serviceDate; }

pythonOne Dark Pro

1 def partition_key(agency_id: str, day: str) -> str:
2     return f"{agency_id}:{day}"

typescriptOne Dark Pro

1 export function isStalePing(tsMs: number, nowMs: number, maxAgeMs = 30000): boolean {
2   return nowMs - tsMs > maxAgeMs;
3 }

Numbers to say aloud

Ingest: 800/s avg, ~2.4k/s peak (8k buses / 10s).
Arrival reads: 50k RPS peak if 5M riders poll every 60s at rush (use push to cut this).
ETA target: p95 error < 60s when live; widen to schedule band when ping age > 45s.

Why interviewers care

a Bus Tracking & Route Optimization System interviews reward crisp scope, explicit trade-offs, and failure stories—not generic microservice diagrams.

Interview checkpoint

Name one failure story for Problem Statement: Public Bus Tracking Platform that proves you understand real outages, not happy-path diagrams.

Key Highlights

•problem focus for sec-01
•State CAP split: AP live, CP schedule/alerts
•Quantify ingest before storage choices

1	def partition_key(agency_id: str, day: str) -> str:
2	return f"{agency_id}:{day}"

1	export function isStalePing(tsMs: number, nowMs: number, maxAgeMs = 30000): boolean {
2	return nowMs - tsMs > maxAgeMs;
3	}