TCP vs UDP: What's Actually Different and When It Matters

TCP vs UDP: What's Actually Different and When It Matters

Two protocols power most of what happens on the internet. Knowing how they differ isn't just theory, it directly affects how you configure proxies, optimize network performance, and choose the right tool for a given task.

TCP in Brief

Connection-oriented. Before any data moves, a handshake establishes the connection and sets transmission parameters. From there, TCP tracks every packet, handles retransmission if something goes missing, and ensures data arrives in order and without errors.

That reliability comes at a cost: overhead. Each acknowledgment, sequence check, and retransmission takes time. For web browsing, email, file transfers, and anything where data integrity matters more than raw speed, that is worth it.

UDP in Brief

Connectionless. Packets go out without setup, without acknowledgment, without guaranteed order. Some arrive, some might not. The protocol doesn't care.

What it lacks in reliability it makes up in speed. No handshake, no congestion control, no waiting for confirmations. For live video, VoIP, online gaming, and real-time applications where a dropped packet is preferable to a delay, UDP is the right call.

Core Differences

Common Ports

TCP ports you'll run into regularly:

• 80 HTTP
• 443 HTTPS
• 21 FTP

UDP ports worth knowing:

• 53 DNS (fast lookups, low latency)
• 5060 VoIP
• 123 NTP (clock synchronization)

How Firewalls and NAT Handle Each

TCP is easier for firewalls to manage. Stateful inspection tracks each session, and return packets are allowed once a connection is established.

UDP is trickier. No handshake means NAT has to guess at session state, which leads to dropped or blocked traffic more often. UDP often needs explicit firewall rules to pass through reliably.

Useful diagnostic tools: traceroute in TCP or UDP mode shows path and packet loss per protocol; nmap identifies which ports are open on a given server.

Protocol Extensions Worth Knowing

TCP has accumulated useful extensions over time:

• Window scaling: increases throughput on high-speed networks
• SACK (Selective Acknowledgments): pinpoints lost packets precisely, reduces unnecessary retransmissions
• Multipath TCP (MPTCP): uses multiple paths simultaneously, useful for fault tolerance in proxy setups

On the UDP side:

• QUIC: adds encryption, error correction, and congestion control on top of UDP. It's what HTTP/3 runs on
• SCTP: combines UDP's low latency with TCP reliability

In Proxy Deployments

Protocol choice has a direct impact on proxy behavior. SOCKS5 runs over TCP by default, it's stable, ordered, suitable for most tasks. For latency sensitive work, UDP-based extensions like QUIC reduce overhead and improve response times.

When fine-tuning matters: on the TCP side, buffer sizing and TCP Fast Open cut down handshake overhead on repeated connections. On the UDP side, jitter buffering handles variable packet arrival cleanly.

For testing, iperf3 covers throughput for both protocols. Wireshark handles deeper packet-level analysis.