For every networking topic, trace one message from sender to receiver through the relevant layers and protocols. State what service each layer provides, what addressing or control information it adds along the way, and how errors, congestion, routing, or access are actually handled. Layering exists because each level solves a narrower communication problem while relying on the services the level below it already provides — this narrowing of scope at each layer is what makes the whole system manageable.
A. Network Basics and Layering
A computer network connects devices so they can exchange data and share services, and networks can be classified by scope (LAN, MAN, WAN), ownership, topology, and architecture. Circuit switching reserves a dedicated path and its resources for an entire communication session; packet switching instead divides data into packets that share network resources dynamically, without any single session monopolizing a dedicated path. Example: traditional telephone circuits illustrate reserved-connection circuit switching, while the modern Internet mainly uses packet switching. Frequency-division multiplexing separates multiple channels by frequency; time-division multiplexing instead allocates distinct time slots to different channels sharing the same physical medium.
Layering divides network communication into manageable, independent services, each using the layer below it and serving the layer above it. OSI is a seven-layer reference model; Internet networking in practice is more commonly described using a four- or five-layer TCP/IP stack. Application-layer protocols include HTTP (web transfer), FTP (file transfer), email protocols, and DNS (naming). Example: when a browser requests a web page, the application data is progressively wrapped with transport, network, and link-layer information as it moves down the sender’s protocol stack, and this wrapping is correspondingly unwrapped, layer by layer, at the receiver — this process is called encapsulation and de-encapsulation. Memorizing what service each layer actually provides, rather than just memorizing layer names in order, is what lets you correctly place any given protocol into its proper layer on an exam.
B. Transport and Network Layers
The transport layer provides end-to-end communication directly between application processes, using ports to allow multiple applications on the same machine to be multiplexed over the same network connection. UDP is connectionless and lightweight, sending data with minimal overhead and no delivery guarantee; TCP provides a reliable, ordered byte stream, using connection setup, sequence numbers and acknowledgments to confirm delivery, retransmission of lost data, and congestion-control mechanisms to avoid overwhelming the network. Example: live voice communication may prefer low delay and can tolerate some data loss, favoring UDP, while a file transfer needs guaranteed reliable delivery of every byte, favoring TCP — this single trade-off between latency and reliability explains why both protocols exist side by side rather than one replacing the other.
The network layer delivers packets across multiple interconnected networks, with the Internet Protocol (IP) providing addressing and forwarding. IPv4 uses 32-bit addresses, and while classful addressing is historically important background, modern routing mainly uses CIDR (classless addressing) instead. ARP resolves an IPv4 address to its corresponding local link-layer address for communication within a LAN; ICMP carries control and error information (such as the messages behind the “ping” utility); NAT translates between address spaces, commonly letting many private devices share one public IPv4 address for outbound connections; IPv6 substantially expands the available address space and changes several header and neighbor-discovery mechanisms compared to IPv4. Example: a home router commonly uses NAT so that many private devices on the home network can all share a single public IPv4 address when connecting outward to the Internet. Routing is the process of selecting the actual paths packets take across this interconnected network of networks.
C. WAN Technologies, Physical/Link Layer, and LANs
Several technologies illustrate the historical evolution of wide-area networking: X.25 emphasized reliable virtual circuits built over older, less reliable physical links; Frame Relay simplified packet switching relative to X.25’s heavier overhead; ATM used fixed-size cells specifically to support predictable, consistent switching behaviour; MPLS forwards traffic using labels to steer it efficiently through provider networks. These differ in era, protocol overhead, service model, and current relevance — several are now mainly legacy technologies, but they remain worth understanding because they teach the underlying design trade-offs of wide-area networking, independent of which specific technology happens to be currently fashionable.
The physical layer moves raw signals or bits across a medium; the link layer delivers frames over a single local link and handles local addressing and shared-medium access. Error detection commonly uses parity bits, checksums, or CRC (cyclic redundancy check); ARQ (automatic repeat request) uses feedback and retransmission to recover from detected errors. Bridges and switches forward frames using link-layer (MAC) addresses, learning which addresses are reachable through which physical port over time; hubs, by contrast, simply repeat incoming signals out to every other port without any such learning or filtering. Example: a switch learns which MAC addresses are reachable on each of its ports and forwards frames selectively, while a hub repeats incoming bits out to all ports indiscriminately, regardless of the intended destination. Multiple-access methods coordinate how several devices share a single physical medium without their transmissions destructively colliding.
D. Network Security, Overlays, and Content Distribution
Modern networks add logical systems on top of basic physical connectivity to improve security, naming, content distribution, and decentralized communication. An overlay network creates logical links running over the top of another underlying network, essentially building a new virtual network topology on top of existing infrastructure. A CDN (content delivery network) places copies of content physically near end users, in geographically distributed nodes, so that not every user request has to travel all the way back to one single origin server. Example: a video streaming service can cache its popular files in CDN nodes distributed close to users worldwide, dramatically cutting the distance data has to travel and improving load times. P2P (peer-to-peer) systems let ordinary peers act as both clients and servers simultaneously, rather than relying on one central server for everything; distributed hash tables (DHTs) map keys to specific nodes in a decentralized way, enabling P2P systems to locate data without any central directory. Common network attacks include eavesdropping (secretly intercepting communication), spoofing (falsifying identity or address information), denial-of-service attacks (deliberately overwhelming a system so legitimate users cannot use it), and malware-assisted compromise. A strong security answer should always distinguish confidentiality, authentication, integrity, and availability as separate, specific security properties, rather than treating “security” as one undifferentiated concept.
Computer Networks key takeaways: tracing one message through the full layered stack, from sender to receiver, is the most reliable way to understand any specific networking topic. Circuit switching reserves resources for a session; packet switching shares resources dynamically across many sessions at once. UDP and TCP trade off latency against reliability, and this single trade-off explains why both protocols persist side by side. IP addressing, ARP, ICMP, and NAT together solve the distinct problems of addressing, local resolution, error signalling, and address-space sharing, respectively. Switches learn and forward selectively using MAC addresses; hubs simply repeat everything indiscriminately. CDNs, P2P systems, and DHTs are all logical structures layered on top of ordinary connectivity to solve distribution and decentralization problems that basic networking alone does not address.