Design a Long-Distance Bus Ticketing Platform - System Design | WinJob

Problem Statement: Long-Distance Bus Marketplace

Design a RedBus / FlixBus / Wanderu-class bus marketplace: multi-operator aggregation, interactive seat maps, dynamic pricing, festival surge fairness, operator webhooks, and QR boarding passes. This section covers problem statement: long-distance bus marketplace in the understanding phase.

Operational detail

Assume 40M registered travelers, 1.2M bookings/day, ~18K concurrent departures, festival windows up to 45K bookings/min on ~200 hot trips. Search averages ~3.5K RPS with 12× morning peak; holds expire in 8 minutes. Hot seat counts live in Redis shards keyed by trip_id, authoritative bookings in PostgreSQL with optimistic inventory_version. Operator GPS ingest ~900 events/s when enabled. CAP: CP for confirmed seats, AP for search ranking with price_as_of timestamps.

Failure and edge cases

Double booking without version check → 409 INVENTORY_STALE
Festival thundering herd → per-user token bucket + trip-scoped queue
Operator webhook delay → booking stays PENDING_OPERATOR with visible SLA
Stale seat map UI → ETag on layout blob; server rejects unknown seat_ids
PSP success but operator reject → saga issues auto-refund + releases seats

Interview checkpoints

javaOne Dark Pro

1 public record HoldToken(String tripId, int inventoryVersion, Instant expiresAt) {}

pythonOne Dark Pro

1 def stop_ranges_overlap(a_from: int, a_to: int, b_from: int, b_to: int) -> bool:
2     return max(a_from, b_from) < min(a_to, b_to)  # order 1

typescriptOne Dark Pro

1 export interface Booking { id: string; tripId: string; seats: SeatLease[]; state: "HELD" | "CONFIRMED"; }
2 export function holdKey(tripId: string): string { return `hold:${tripId}`; }

Numbers to say aloud

Festival: 45K bookings/min ≈ 750 holds/s on narrow trip shards — size pools before debating Kafka.
Stop-pair index precomputed per trip; allocator is O(log n) shard lookup, not full scan.
Live bus ETA: p95 error < 120s when GPS fresh; schedule-only band when ping age > 90s.

Why interviewers care

a Long-Distance Bus Ticketing Platform interviews reward crisp scope, explicit trade-offs, and failure stories—not generic microservice diagrams.

Interview checkpoint

Name one failure story for Problem Statement: Long-Distance Bus Marketplace that proves you understand real outages, not happy-path diagrams.

Key Highlights

•focus for sec-01
•CP inventory vs AP search cache
•Quantify festival shard load before queues

1	def stop_ranges_overlap(a_from: int, a_to: int, b_from: int, b_to: int) -> bool:
2	return max(a_from, b_from) < min(a_to, b_to) # order 1

1	export interface Booking { id: string; tripId: string; seats: SeatLease[]; state: "HELD" \| "CONFIRMED"; }
2	export function holdKey(tripId: string): string { return `hold:${tripId}`; }