TCP vs UDP

TCP vs UDP

Why This Matters

TCP is like sending a registered letter—you get confirmation it arrived, and it arrives in order. UDP is like shouting across a crowded room—you hope someone hears you, but you have no guarantee.

This distinction matters because it determines how your system handles failures. TCP guarantees delivery, but that guarantee has overhead (handshakes, acknowledgments, retries). UDP has no overhead, but packets can be lost, duplicated, or arrive out of order.

In interviews, when someone asks "How would you design a real-time multiplayer game?", they're testing whether you understand this trade-off. Do you use TCP for reliability, or UDP for speed? The answer is usually "both"—TCP for critical events, UDP for frequent updates.

What Engineers Usually Get Wrong

Most engineers think "TCP is reliable, so use TCP for everything." But TCP's reliability comes with overhead. Every TCP connection requires a three-way handshake (one round trip = 50-100ms). For short-lived connections, this overhead is significant. Also, TCP has flow control and congestion control, which can slow down data transfer when the network is congested.

Engineers also think "UDP is unreliable, so never use it." But UDP is perfect for applications where speed matters more than perfect delivery. Video streaming, online gaming, DNS queries—these all use UDP because a lost packet is less important than low latency.

How This Breaks Systems in the Real World

A microservice was making HTTP requests to another service. Each request opened a new TCP connection, sent data, then closed the connection. Under normal load, this worked fine. But during a traffic spike, the service tried to open 10,000 connections per second. Each connection required a three-way handshake (one round trip = 50ms). The handshake queue filled up. New connections timed out. The service became unresponsive.

The fix? Use HTTP keep-alive and connection pooling. Reuse TCP connections instead of creating new ones for each request. But the real lesson is: TCP connections have overhead. Opening and closing connections is expensive. Reuse them.

Another story: A real-time multiplayer game was using TCP for all network communication. The game felt laggy. Investigation showed that TCP's congestion control was slowing down game state updates. When the network was congested, TCP would reduce the sending rate, causing the game to feel unresponsive. The fix? Use UDP for game state updates (position, movement), and TCP only for critical events (player actions, chat). This gave the game low latency while maintaining reliability for important events.

TCP (Transmission Control Protocol)

Characteristics:

• Connection-oriented: Establishes connection before data transfer
• Reliable: Guarantees delivery, ordered delivery
• Flow control: Adjusts sending rate based on receiver capacity
• Congestion control: Adapts to network conditions
• Full-duplex: Bidirectional communication

Header size: 20-60 bytes

Use cases: Web browsing (HTTP), email (SMTP), file transfer (FTP), database connections

TCP Example: HTTP Request

1 

UDP (User Datagram Protocol)

Characteristics:

• Connectionless: No connection establishment
• Unreliable: No delivery guarantee, no ordering
• No flow control: Sends at maximum rate
• No congestion control: Doesn't adapt to network
• Lightweight: Minimal overhead

Header size: 8 bytes

Use cases: DNS queries, video streaming, online gaming, VoIP, DHCP

UDP Example: DNS Query

1 

Comparison

Failure Stories You'll Recognize

The Connection Storm: A service was creating new TCP connections for every request. During a traffic spike, it tried to create 10,000 connections per second. The three-way handshake queue filled up. New connections timed out. The service appeared down. The fix? Use connection pooling.

The Laggy Game: A multiplayer game used TCP for all communication. The game felt laggy because TCP's congestion control slowed down updates during network congestion. The fix? Use UDP for frequent updates, TCP for critical events.

The Lost DNS Queries: A service was making DNS queries over TCP instead of UDP. DNS queries are small and need to be fast. TCP's overhead made DNS lookups slow. The fix? Use UDP for DNS (the standard).

What Interviewers Are Really Testing

They want to hear you talk about the trade-off between reliability and latency. Junior engineers say "use TCP for everything." Senior engineers say "TCP for reliability when you need it, UDP for speed when you can tolerate loss."

When they ask "How would you design a real-time system?", they're testing:

• Do you understand that TCP's reliability has overhead?
• Do you know when UDP is appropriate?
• Can you design a hybrid approach that uses both?

Interview Questions

Beginner

Q: What are the main differences between TCP and UDP?

A:

TCP:

• Connection-oriented, reliable, ordered delivery
• Higher overhead, slower
• Use when data integrity is critical

UDP:

• Connectionless, unreliable, no ordering
• Lower overhead, faster
• Use when speed is critical and some loss is acceptable

Example: HTTP uses TCP (need reliable delivery). DNS uses UDP (speed matters, can retry).

Intermediate

Q: When would you choose UDP over TCP? Give examples.

A:

Choose UDP when:

1. Real-time applications: Video streaming, gaming (latency critical)
2. Simple request-response: DNS queries (can retry if lost)
3. Broadcast/multicast: One-to-many communication
4. High throughput: When you need maximum speed

Examples:

• DNS: UDP for queries (fast, can retry)
• Video streaming: UDP for live video (some packet loss acceptable)
• Online gaming: UDP for game state (old updates irrelevant)
• VoIP: UDP for voice (low latency > perfect delivery)

Trade-off: Accept some packet loss for lower latency.

Senior

Q: Design a real-time multiplayer game network protocol. How do you balance reliability and latency? When would you use TCP vs UDP?

A:

Design: Hybrid Approach

1 

Strategy:

• UDP for frequent updates: Position, movement (30-60 updates/sec)
• TCP for critical events: Player actions, chat, game events
• Sequence numbers: Detect missing UDP packets
• Selective reliability: Retry only important packets via TCP
• Compression: Reduce UDP packet size

Why this works:

• Low latency for game state (UDP)
• Reliability for critical events (TCP)
• Best of both worlds