Computer Networks: Architecture, Protocols, Operation, and Security

A computer network is a system of devices, links, and software protocols that enables hosts to exchange data across local, metropolitan, wide-area, and global environments. The Internet is the largest practical example: it uses packet-based communication, standardized addressing, routing, and layered protocols to move data between heterogeneous systems .

At the core of modern networking is packet switching, where application data is segmented into packets, each packet is forwarded through intermediate devices, and the receiver reassembles the original message. This design allows links to be shared efficiently among many conversations, unlike circuit-switched systems that reserve a fixed path for an entire session .

Networks are usually studied through layering. Layered models separate concerns: physical signaling, local delivery, internetwork routing, transport reliability, and application semantics. The OSI model is a conceptual seven-layer reference model, while the TCP/IP model is the practical foundation of the Internet protocol suite .

Key learning outcomes for this section:

Footnotes

1. RFC 1122: Requirements for Internet Hosts — Communication Layers - Foundational specification describing Internet host communication layers and TCP/IP requirements. ↩

2. Internet Society: A Brief History of the Internet - Historical overview of packet switching, ARPANET, and Internet development. ↩

3. ISO/IEC 7498-1: Open Systems Interconnection Basic Reference Model - International standard defining the OSI reference model. ↩

Network Components and Topologies

A network is built from endpoint devices, intermediary devices, links, and protocols. A host may be a laptop, server, phone, printer, virtual machine, or container. A switch connects devices inside a local area network, while a router connects separate IP networks and chooses next hops for packets .

A LAN usually operates over Ethernet or Wi-Fi. A WAN connects remote sites through carrier links, leased lines, VPNs, or public Internet paths. A topology describes how nodes are arranged and how traffic flows.

Common topology patterns:

Footnotes

1. Cisco: What Is a Network Switch? - Vendor-neutral educational explanation of switching and network device roles. ↩

Addressing: MAC, IP, Ports, and Names

Networking uses several kinds of identifiers because different layers solve different problems. A MAC address identifies an interface on a local link. An IP address identifies a host or interface within an IP network. A port number identifies an application conversation on a host.

IPv4 uses 32-bit addresses, commonly written in dotted-decimal form such as 192.0.2.10; IPv6 uses 128-bit addresses, commonly written in hexadecimal colon-separated notation such as 2001:db8::10 . IPv6 was designed with a vastly larger address space and improved support for modern Internet growth .

A subnet groups addresses under a common prefix. For example, 192.0.2.0/24 indicates that the first 24 bits are the network prefix, leaving 8 bits for host addresses. Classless Inter-Domain Routing, or CIDR, allows flexible address allocation and route aggregation .

The DNS translates names such as example.com into addresses that machines can route to. DNS is hierarchical and distributed, using zones, authoritative servers, recursive resolvers, and cached records to scale globally .

Footnotes

1. RFC 791: Internet Protocol - Original IPv4 specification defining IP datagrams, addressing, and fragmentation. ↩

2. RFC 8200: Internet Protocol, Version 6 Specification - Current IPv6 specification describing IPv6 packet format and addressing architecture. ↩

3. RFC 4632: Classless Inter-domain Routing - Specification and discussion of CIDR addressing and route aggregation. ↩

4. RFC 1034: Domain Names — Concepts and Facilities - Core DNS architecture specification explaining the distributed domain name system. ↩

Switching, Routing, and Forwarding

A switch operates primarily at the data-link layer. It forwards frames within the same LAN. Ethernet switches maintain forwarding tables that associate MAC addresses with switch ports, reducing unnecessary traffic compared with older shared-medium designs .

A router operates at the network layer. It forwards packets between networks based on destination IP prefixes. Routers use routing tables populated by static configuration, directly connected networks, or dynamic routing protocols.

A routing table usually contains:

A key distinction is that switching is local and link-specific, while routing is logical and network-wide. Switches ask, “Which port reaches this MAC address?” Routers ask, “Which next hop reaches this IP prefix?”

Footnotes

1. Cisco: What Is a Network Switch? - Vendor-neutral educational explanation of switching and network device roles. ↩

Transport Protocols: TCP, UDP, and QUIC

The transport layer provides process-to-process communication. TCP provides reliable byte-stream delivery, retransmission, flow control, congestion control, and ordered data delivery . UDP provides datagram delivery with lower overhead but without TCP-style reliability or ordering guarantees .

TCP is appropriate when correctness and ordered delivery matter, such as web transfers, file downloads, email delivery, and remote administration. UDP is appropriate when applications can tolerate loss or implement their own recovery, such as voice, video, gaming, telemetry, and DNS queries.

QUIC is a modern transport protocol that runs over UDP and integrates security, multiplexing, and connection migration. HTTP/3 uses QUIC rather than TCP, helping reduce connection setup costs and avoiding some head-of-line blocking issues associated with TCP-based multiplexing .

Transport-layer comparison:

The simplified throughput relationship often used in performance analysis is:

$$\text{Throughput} \approx \frac{\text{Data transferred}}{\text{Transfer time}}$$

Latency-sensitive applications care not only about throughput, but also about jitter, packet loss, and round-trip time.

Footnotes

1. RFC 9293: Transmission Control Protocol - Modern TCP specification describing reliable ordered transport behavior. ↩

2. RFC 768: User Datagram Protocol - UDP specification defining connectionless datagram transport. ↩

3. RFC 9000: QUIC — A UDP-Based Multiplexed and Secure Transport - QUIC transport protocol standard. ↩

Application Protocols and Services

Application protocols define the meaning of exchanged messages. HTTP defines semantics for web requests and responses, including methods, status codes, headers, and representation metadata . DNS defines distributed name resolution. SMTP, IMAP, and related protocols support email. SSH supports secure remote administration. DHCP helps hosts obtain IP configuration automatically.

A client-server model is common: browsers request pages from web servers, mail clients retrieve messages from mail servers, and applications query databases through network APIs. A peer-to-peer model distributes service roles among participating nodes.

Typical protocol stack for loading a secure website:

TLS protects data in transit and authenticates servers using certificates. TLS 1.3 simplified older handshake designs and removed many obsolete cryptographic options .

Footnotes

1. RFC 9110: HTTP Semantics - Current HTTP semantics specification for methods, status codes, and message meaning. ↩

2. RFC 8446: The Transport Layer Security Protocol Version 1.3 - TLS 1.3 specification for secure network communication. ↩

Network Security Fundamentals

Network security combines prevention, detection, response, and recovery. A firewall enforces traffic policy between zones or hosts. Network segmentation limits the blast radius of failures and attacks. A VPN protects traffic across untrusted networks.

Security objectives are often summarized as confidentiality, integrity, and availability:

Firewalls are effective when deployed as part of a broader architecture that includes least privilege, patching, authentication, logging, and incident response . Encryption protects data in transit, but it does not replace endpoint security or sound access control.

Footnotes

1. NIST SP 800-41 Rev. 1: Guidelines on Firewalls and Firewall Policy - NIST guidance on firewall technologies, policy, and network security architecture. ↩

Design Principles for Reliable Networks

Good network design balances scalability, performance, security, and operability. A scalable design uses hierarchical addressing, route summarization, modular topology, and predictable failure domains. A secure design applies least privilege, segmentation, encryption, centralized logging, and tested incident response. An operable design includes documentation, monitoring, configuration management, and change control.

Important design metrics include:

A useful availability approximation is:

$$\text{Availability} = \frac{\text{Uptime}}{\text{Uptime} + \text{Downtime}}$$

For example, a service that is down for 8.76 hours per year has approximately $99.9\%$ annual availability. Higher availability usually requires redundancy, monitoring, tested failover, and disciplined operations rather than only faster hardware.