Design an Auction Platform — Real-time Bidding System

Problem Statement

Design an online auction platform where users can:

• Create auctions with a start time, end time, and minimum bid
• Place bids on active auctions
• View a near real-time leaderboard of top bids
• Determine the winning bid when the auction closes

The system should handle high concurrency (thousands of bids per second on popular auctions).

Requirements

Functional:

• Create auction (with start/end time, min bid, item details)
• Place a bid (must be higher than current highest)
• View current top bids / leaderboard
• Determine winner at auction close
• Notify winner and seller

Non-Functional:

• Low latency bid placement (< 100ms)
• Consistent — no two users should "win" simultaneously
• Support 10K+ concurrent bidders on a single auction
• Near real-time leaderboard updates

What the Interviewer Expects

1. Database schema — auctions table, bids table, users table. Discuss indexing on (auction_id, amount) for fast lookups.
2. Bid placement flow — validate bid > current max, insert atomically. Discuss race conditions.
3. Caching layer — Redis sorted sets for real-time leaderboard. Current highest bid cached for fast validation.
4. Consistency vs speed trade-off — Redis for reads (fast leaderboard), DB as source of truth for the final winner.
5. Auction close logic — scheduled job or event-driven. How to atomically determine winner.

Follow-ups

1. How do you handle bid edits? (Allow multiple bids vs edit-in-place — trade-offs)
2. How would you support a near real-time leaderboard? (Redis sorted sets with ZADD)
3. How do you identify the winning bid reliably? (Can't rely solely on Redis — need DB confirmation with locking)
4. What happens if the auction close event fails? (Idempotency, retry logic)
5. How would you prevent shill bidding (fake bids to drive up price)?