How the Internet Actually Works: Packets, Protocols, and the Hidden Infrastructure of Modern Technology

01 The Packet-Switching Origins

The internet began as a radical idea: instead of dedicating a single circuit to each communication, break data into small chunks called packets and route them independently across a shared network. This concept, first articulated by Paul Baran at the RAND Corporation in the early 1960s and independently developed by Donald Davies at the National Physical Laboratory in the UK, became the foundation of packet switching. The ARPANET, launched in 1969 with four nodes at UCLA, Stanford Research Institute, UC Santa Barbara, and the University of Utah, proved that distributed packet-switched networks could survive partial failures — a design born from Cold War concerns about command-and-control resilience.

Each packet carries a header containing source and destination addresses, a sequence number, and error-checking data. When a packet arrives at a router, the router examines the destination address and forwards the packet along the best available path. If a link fails, packets are rerouted automatically. This self-healing property is what makes the internet robust: there is no single point of failure in a well-designed packet-switched network, because every packet can take a different route to the same destination.

02 The TCP/IP Protocol Stack

The internet runs on a layered protocol stack known as TCP/IP, named after its two most important protocols: the Transmission Control Protocol and the Internet Protocol. The stack is typically described in four layers. The Link Layer handles physical transmission over Ethernet cables, fiber optics, or wireless radio. The Internet Layer routes packets across networks using IP addresses. The Transport Layer manages end-to-end communication — TCP provides reliable, ordered delivery, while UDP offers fast, connectionless transmission. The Application Layer contains the protocols users interact with directly: HTTP for web pages, SMTP for email, DNS for name resolution, and dozens of others.

This layering is not merely academic. Each layer depends only on the one below it, meaning a web application using HTTP does not need to know whether the underlying physical connection is fiber optic, Wi-Fi, or a cellular LTE link. This abstraction is what allows the same web browser to work identically on a desktop computer, a smartphone, or a satellite terminal. The protocol stack's separation of concerns is arguably the most successful architectural decision in computing history.

03 DNS: The Phonebook of the Internet

Humans remember names; computers process numbers. The Domain Name System bridges this gap by translating human-readable domain names like sailorbob.org into machine-readable IP addresses like 74.208.95.219. DNS is a hierarchical, distributed database replicated across millions of servers worldwide. When a user types a URL, the browser queries a recursive resolver, which in turn queries root name servers, top-level domain servers, and authoritative name servers to find the answer. This typically takes milliseconds.

The system's distributed design means no single server holds the entire database. Root servers know only about top-level domains (.com, .org, .net). TLD servers know about individual registrants. Authoritative servers know the specific records for their domain. This hierarchy makes DNS both fast and resilient — if one server fails, others can answer from cache or redirect to alternatives. The DNSSEC extension adds cryptographic signatures to prevent spoofing, though adoption remains incomplete.

04 Routing and BGP: The Postal System of Networks

The internet has no central map. Instead, autonomous systems — the large networks operated by ISPs, cloud providers, and enterprises — exchange routing information using the Border Gateway Protocol. BGP is the protocol that holds the internet together, allowing over 70,000 autonomous systems to advertise which IP address blocks they can reach. When a packet leaves a user's device, it travels through a series of routers, each making a local forwarding decision based on its routing table.

BGP is powerful but fragile. A misconfigured router can announce that it owns IP blocks it does not actually control, causing traffic to flow to the wrong destination — a phenomenon known as route hijacking. Several major outages, including the 2019 brief disruption of much of the global internet, have been traced to BGP errors. Despite its criticality, BGP was not designed with built-in security, and efforts like Resource Public Key Infrastructure (RPKI) are gradually adding cryptographic validation to the routing system.

05 The Physical Infrastructure: Cables, Towers, and Data Centers

The internet feels wireless, but it is overwhelmingly physical. Over 95 percent of intercontinental data traffic travels through submarine fiber optic cables laid across ocean floors — over 530 active cables spanning more than 1.4 million kilometers. These cables, some as thick as a garden hose, carry terabits of data per second using pulses of laser light through strands of glass thinner than a human hair. On land, fiber backbone networks connect cities, while cell towers and Wi-Fi access points provide the last-mile wireless connection to mobile devices.

Data centers house the servers that run websites, applications, and cloud services. A single hyperscale data center can contain tens of thousands of servers consuming tens of megawatts of electricity. Companies like Amazon, Google, Microsoft, and Meta operate hundreds of these facilities worldwide, strategically located near cheap electricity and fiber backbones. Content Delivery Networks (CDNs) like Cloudflare and Fastly cache content at edge locations close to users, reducing latency by serving data from nearby servers rather than distant origin data centers.

06 How Cellular Phones Connect

When a smartphone loads a webpage, the data does not travel through a wire — it hops from the phone to a nearby cell tower over a radio link. The tower connects to a base station controller, which connects to a mobile switching center, which connects to the core internet backbone. Modern cellular networks use digital modulation schemes that pack more data into limited radio spectrum. A 5G connection can theoretically reach 10 Gbps, though real-world speeds are far lower due to signal interference, tower congestion, and distance.

The cellular network is itself a layered system. The radio access network handles the air interface. The core network manages mobility, authentication, and routing. When a user moves between cell towers, the network performs a handoff — transferring the connection without dropping it. This seamless mobility, combined with the internet's packet-switching architecture, is what makes it possible to stream video while riding in a moving car.

07 Emerging Challenges: Bandwidth, Security, and Decentralization

The internet faces growing pressures. Video streaming already accounts for over 60 percent of global internet traffic, and AI workloads are adding new demands. The shift to encrypted connections (HTTPS) has improved privacy but complicated content filtering and caching. Cybersecurity threats — from distributed denial-of-service attacks to BGP hijacking to DNS spoofing — exploit the protocol stack's original assumption of trusted participants, an assumption that no longer holds in a network connecting billions of devices.

Efforts to decentralize the internet — through mesh networking, peer-to-peer protocols, and edge computing — aim to reduce dependence on large cloud providers. Meanwhile, satellite internet services like Starlink extend connectivity to remote regions that terrestrial infrastructure cannot reach. The fundamental architecture, however, remains unchanged: packets routed across autonomous networks, bound together by shared protocols designed more than half a century ago.