Design Instagram Auction System

Problem Statement

Create a system that enables content creators and retailers to host live shopping events where they showcase products through video streaming while viewers can browse items, ask questions, and make instant purchases. The platform should handle concurrent live streams, real-time interactions, inventory management, and seamless checkout experiences. Consider that popular influencers might attract hundreds of thousands of concurrent viewers, while the average stream might have only a few dozen participants. The system must balance low-latency interactions for engaged shoppers with cost-effective infrastructure for the long tail of smaller events.

Key Requirements

Functional

• Live video streaming -- creators broadcast live video showcasing products with the ability to highlight specific items during the stream
• Real-time product catalog -- dynamically display products featured in the stream with pricing, descriptions, and availability
• Interactive features -- viewers can ask questions via chat, react with emojis, and receive notifications when featured items go on sale
• Instant checkout -- users can purchase products without leaving the stream, with cart and payment processing
• Stream discovery -- users can browse upcoming and live shopping events, with personalized recommendations
• Analytics dashboard -- creators see real-time metrics on viewership, engagement, and sales performance

Non-Functional

• Scalability -- support 10,000+ concurrent streams with individual streams handling 1K to 500K viewers; process 100K+ purchases per hour during peak events
• Reliability -- ensure 99.9% uptime for checkout flow; graceful degradation for non-critical features during overload
• Latency -- video delay under 3 seconds from broadcaster to viewers; chat messages delivered within 500ms; inventory updates reflected within 2 seconds
• Consistency -- strong consistency for inventory and payment transactions; eventual consistency acceptable for view counts and social features

What Interviewers Focus On

Based on real interview experiences, these are the areas interviewers probe most deeply:

1. Video Streaming Architecture

The video delivery system is critical since poor quality directly impacts sales conversion. Interviewers want to see how you handle varying audience sizes and geographic distribution.

Hints to consider:

• Explore adaptive bitrate streaming protocols like HLS or DASH to accommodate different network conditions
• Consider a CDN strategy that scales dynamically based on viewer count and uses edge locations for regional optimization
• Discuss tradeoffs between WebRTC for ultra-low latency versus traditional streaming for scalability
• Think about how to handle the "thundering herd" problem when a celebrity announces a stream on social media

2. Inventory Management Under High Concurrency

Selling limited inventory items during live events creates race conditions where thousands might attempt to purchase the same product simultaneously. This is a core technical challenge.

Hints to consider:

• Design a reservation system that temporarily holds inventory during checkout to prevent overselling
• Consider optimistic locking strategies and how to handle conflicts gracefully
• Discuss queue-based approaches for managing purchase requests during viral moments
• Think about how to display real-time stock counts without creating database hotspots

3. Real-Time Interaction Synchronization

Keeping chat, reactions, and product highlights synchronized across potentially hundreds of thousands of viewers requires careful design.

Hints to consider:

• Explore pub-sub architectures using WebSockets or Server-Sent Events for pushing updates to clients
• Consider fan-out strategies and whether to use push or pull models for different feature categories
• Discuss how to partition chat rooms to prevent a single stream's chat from overwhelming infrastructure
• Think about client-side buffering and reconciliation to handle temporary network issues

4. Data Consistency Models

Different parts of the system have different consistency requirements that impact both user experience and engineering complexity.

Hints to consider:

• Payment and inventory transactions require strong consistency to prevent financial discrepancies
• View counts, likes, and non-critical analytics can use eventual consistency for better performance
• Consider using CQRS to separate read and write paths with different consistency guarantees
• Discuss how to handle distributed transactions across inventory, payment, and order services

Suggested Approach

Step 1: Clarify Requirements

Start by establishing the scope and scale parameters. Confirm the expected ratio of viewers to purchasers (typically 5-10% conversion). Ask about the geographic distribution of users and whether you need to support international payments and shipping. Clarify whether creators can pre-schedule products or if they must add them during the live stream. Determine if the system needs to support recorded playback of past shopping events or only live streams. Understand the business model -- does the platform take a transaction fee, charge creators for hosting, or both?

Step 2: High-Level Architecture

Sketch the main components: a video ingestion service that receives RTMP streams from creators, transcoding service that converts to multiple bitrates, CDN for distributing video to viewers, and origin servers for dynamic content. Include a product catalog service managing inventory and pricing, a real-time messaging service handling chat and interactions, and a checkout service coordinating cart, payment processing, and order fulfillment. Add a notification service for alerting users about upcoming streams and featured products. Show data stores including a relational database for transactions, NoSQL for product catalogs and user profiles, and a cache layer for hot data.

Step 3: Deep Dive on Key Area

Focus on the checkout flow and inventory management system. Walk through the journey from a viewer clicking "buy" to order confirmation. Explain how you'll implement a two-phase approach: first, a reservation phase that atomically decrements available inventory and creates a time-limited hold (maybe 5 minutes), then a completion phase after payment succeeds. Discuss using a message queue to process purchase requests asynchronously, preventing overwhelming the database during spikes. Detail how you'll handle reservation expiration and inventory release using TTL mechanisms. Address the edge case where multiple concurrent purchases compete for the last item, explaining your conflict resolution strategy (first-come-first-served with clear user feedback versus a lottery system for fairness).

Step 4: Address Secondary Concerns

Discuss monitoring and observability, emphasizing real-time dashboards for creators showing current viewership and sales velocity. Cover your strategy for handling abusive behavior in chat using rate limiting and content moderation. Address cost optimization by auto-scaling video infrastructure based on actual viewer count rather than provisioning for peak. Explain your approach to testing, particularly chaos engineering for simulating flash sale scenarios. Touch on internationalization concerns including multi-currency support and regional CDN distribution. Mention potential future enhancements like AI-powered product recommendations during streams or augmented reality try-on features.