Computer Networking: A Top-Down Approach - Chapter 1: Introduction

Conceptual Perspectives of the Internet

  • Nuts-and-Bolts View of Internet Infrastructure:

    • Millions of interconnected computing devices, formally termed hosts or end systems, execute network-based applications.

    • Packet switches forward packets (discrete chunks of serialized data); primary switch varieties include routers (prevalent in the network core) and link-layer switches (predominant in access networks).

    • Communication links physically span distances to connect switches and hosts, utilizing diverse transmission media such as coaxial cable, copper twisted pair, optical fiber, terrestrial radio, and satellite channels.

    • Links are characterized by their transmission rate, commonly referred to as link capacity or link bandwidth, measured in bits per second (bps\text{bps}).

    • The Internet functions as a "network of networks," interconnecting diverse Internet Service Providers (ISPs), ranging from local residential and enterprise access providers to regional ISPs and Tier-1 global ISPs.

    • Embedded and non-traditional devices connected to the Internet include IP-enabled picture frames (e.g., Ceiva), Web-enabled toasters with integrated weather forecasting displays, Internet telephones (VoIP devices), Internet-connected refrigerators, remote media streaming devices (e.g., Slingbox), and networked energy monitors (e.g., Tweet-a-watt).

Overview of Internet edge, access networks, and core ISPs
  • Service-Oriented View of the Internet:

    • Provides a shared telecommunication infrastructure that delivers services to distributed applications, such as the World Wide Web, streaming audio/video, Voice over IP (VoIP), peer-to-peer file sharing, e-commerce, online gaming, and social networks.

    • Exposes an Application Programming Interface (API) to distributed software components, providing software hooks and configuration options that govern how an application running on one end system instructs the Internet infrastructure to deliver data to another end system.

    • Offers alternative service models analogous to traditional postal options (e.g., guaranteed reliable connection-oriented delivery versus best-effort connectionless delivery).

  • Protocols in Computer Networking:    

    • Definition: A protocol defines the syntax, format, and semantic order of messages exchanged between two or more communicating entities, as well as the specific actions taken upon the transmission or receipt of a message, or upon the occurrence of an internal system event.

    • Human Analogy: Human social interactions follow rigid protocols (e.g., asking "Got the time?" requires a greeting, a query, a time response like "2:00", and an acknowledgment).

    • Computer Protocol Interaction: To request a Web resource, a client host transmits a TCP connection request packet; the server responds with a TCP connection response packet; the client then issues an HTTP GET request (e.g., GET http://www.awl.com/kurose-ross); the server processes the request and responds by transmitting the requested file.

    • Standards Organizations:

    • IETF (Internet Engineering Task Force): Develops official Internet standards published as RFCs (Request for Comments).

    • IEEE 802 LAN/MAN Standards Committee: Formulates local and metropolitan area network link standards (e.g., Ethernet and IEEE 802.11 WiFi).

Network Edge: End Systems, Access Networks, and Physical Media

  • Components of the Network Edge:

    • End Systems (Hosts): Machines residing at the periphery of the Internet that run end-user application software; divided into clients (desktop PCs, smartphones, laptops) and servers (powerful compute platforms typically consolidated inside large data centers).

    • Access Networks: Physical networks and links that connect end systems directly to the first router (termed the edge router) on the path to the rest of the network.

    • Two fundamental design metrics of access networks are transmission bandwidth (bps\text{bps}) and channel sharing model (dedicated versus shared capacity).

  • Digital Subscriber Line (DSL):

    • Utilizes pre-existing copper telephone lines to establish a dedicated, point-to-point connection between a residential subscriber and a telephone company Central Office (CO).

    • Employs Frequency-Division Multiplexing (FDM) over the single copper twisted-pair line to separate communication bands:

    • Two-way telephone channel (Voice): 00\text{ to }4 kHz4\,\text{kHz}.

    • Upstream data channel: 44\text{ to }50 kHz50\,\text{kHz}.

    • Downstream data channel: 50 kHz50\,\text{kHz}\text{ to }1 MHz1\,\text{MHz}.

    • Hardware Setup: A residential splitter separates incoming analog signals, steering low frequencies to telephone hardware and high frequencies to a DSL modem. At the CO, a DSL Access Multiplexer (DSLAM) aggregates digital data from thousands of lines into the ISP backbone while redirecting analog voice signals to the traditional voice telephone network.

    • Transmission Characteristics: DSL is asymmetric; downstream transmission rates reach up to 24 Mbps24\,\text{Mbps} (typically <10 Mbps< 10\,\text{Mbps}), while upstream rates reach up to 2.5 Mbps2.5\,\text{Mbps} (typically <1 Mbps< 1\,\text{Mbps}).

Digital Subscriber Line (DSL) access network architecture
  • Home Network Access Architecture:

    • Connects multiple residential end systems (smartphones, PCs, laptops) using a combination of wired and wireless technologies.

    • Connected to an ISP via a cable modem or DSL modem.

    • Integrates an internal local router, firewall, Network Address Translation (NAT) engine, an Ethernet switch (100 Mbps100\,\text{Mbps}), and a wireless local access point (54 Mbps54\,\text{Mbps} 802.11g) often housed inside a single combined physical chassis.

Home network architecture with modem, router, and wireless access point
  • Enterprise Access Networks (Ethernet):

    • Deployed across institutional environments (universities, corporate campuses).

    • End-user workstations and shared institutional servers (mail servers, web servers) link directly to high-speed Ethernet switches.

    • Common transmission capacities operate at 10 Mbps10\,\text{Mbps}, 100 Mbps100\,\text{Mbps}, 1 Gbps1\,\text{Gbps}, and 10 Gbps10\,\text{Gbps}.

    • Ethernet switches interconnect with institutional routers that maintain high-capacity dedicated links into upstream ISP networks.

Enterprise Ethernet access network connecting institutional servers and hosts
  • Wireless Access Networks:

    • Wireless Local Area Networks (WLANs):

    • Mobile end systems communicate with an access point (base station) over a radius of approximately 100 ft100\,\text{ft} (≈30 m\approx 30\,\text{m}).

    • Governed by IEEE 802.11 standards; 802.11b operates at 11 Mbps11\,\text{Mbps}, and 802.11g operates at 54 Mbps54\,\text{Mbps}.

    • Wide-Area Cellular Wireless Access:

    • Provided by cellular service providers through base stations covering wide geographical cells spanning tens of kilometers (10’s of km10\text{'s of km}).

    • Speeds typically range between 1 Mbps1\,\text{Mbps} and 10 Mbps10\,\text{Mbps}.

    • Built on 3G and 4G (Long-Term Evolution - LTE) technologies.

  • Physical Transmission Media:

    • Guided Media: Solid, physical channels guiding electromagnetic waves:

    • Copper Twisted Pair: Inexpensive, ubiquitous medium used in telephone lines and Category 5/6 Ethernet cabling.

    • Coaxial Cable: Concentric copper conductors delivering broadband frequencies for cable television and DOCSIS access networks.

    • Fiber-Optic Cable: Thin glass strands guiding pulses of light; supports extremely high bit rates (tens to hundreds of gigabits per second), exhibits low attenuation over long distances, and is immune to electromagnetic interference.

    • Unguided Media: Open-air channels through which electromagnetic waves radiate freely:

    • Terrestrial Radio Channels: Carry WiFi, cellular signals, and line-of-sight microwave links; susceptible to multipath fading, path loss, and obstruction interference.

    • Satellite Channels: Geostationary or low-Earth orbit transponders relaying radio signals over expansive geographical spans.

Network Core: Switching Paradigms and Architecture

  • Packet Switching:

    • The network core consists of an interconnected mesh of packet switches and communication links.

    • Source hosts fragment long application-layer messages into smaller segments called packets, each containing L bitsL\,\text{bits}.

    • Packets traverse individual communication links at the link's full transmission capacity R bits/secondR\,\text{bits/second}.

    • Store-and-Forward Transmission: A packet switch must receive all L bitsL\,\text{bits} of an arriving packet before it can begin transmitting the first bit onto the outbound link.

    • One-Hop Transmission Delay:

dtrans=LRd_{\text{trans}} = \frac{L}{R}

Packet transmission delay formula
  • Numerical Store-and-Forward Example: If a packet size is L=7.5 MbitsL = 7.5\,\text{Mbits} and the link rate is R=1.5 MbpsR = 1.5\,\text{Mbps}, the transmission delay for one hop is:

dtrans=7.5×106 bits1.5×106 bps=5 secondsd_{\text{trans}} = \frac{7.5 \times 10^6\,\text{bits}}{1.5 \times 10^6\,\text{bps}} = 5\,\text{seconds}

  • For a path of NN identical links of capacity RR connected by N−1N - 1 intermediate store-and-forward switches, the total end-to-end transmission delay (ignoring propagation and processing delays) is:

dend-to-end=N×LRd_{\text{end-to-end}} = N \times \frac{L}{R}

  • Traversing two links (N=2N = 2) with one intermediate switch yields an end-to-end transmission delay of:

dend-to-end=2×LRd_{\text{end-to-end}} = 2 \times \frac{L}{R}

  • Core Routing and Forwarding Functions:

    • Forwarding: A local, hardware-driven operation executed within a single router; moves arriving packets from an input link interface to the appropriate output link interface based on inspecting destination header fields against the router's local forwarding table.

    • Routing: A network-wide control-plane process that executes distributed routing algorithms across routers to determine end-to-end paths for packets and build the forwarding tables.

  • Circuit Switching:

    • Allocates and reserves end-to-end physical resources (buffers, link transmission bandwidth) for the entire duration of an established communication call.

    • Dedicated resources provide guaranteed, constant transmission capacity with zero queueing delay and zero congestion loss once established.

    • If a reserved circuit remains idle (e.g., during conversational pauses in a telephone call), its allocated bandwidth cannot be utilized by other network traffic.

    • Multiplexing Methods in Circuit Switching:

    • Frequency-Division Multiplexing (FDM): The frequency spectrum of a link is partitioned into discrete, non-overlapping channels; each call receives exclusive access to a narrow frequency slice for the entirety of the connection.

    • Time-Division Multiplexing (TDM): Time is divided into recurring frames consisting of a fixed number of synchronized time slots; each connection transmits at the full link speed only during its assigned time slot in every frame.

  • Comparison: Packet Switching versus Circuit Switching:

    • Packet switching is well suited for bursty data traffic; it promotes efficient statistical resource sharing without requiring call setup overhead.

    • High arrival rates in packet switching can lead to link buffer congestion, generating variable packet delays and packet loss.

    • Providing circuit-like performance (bandwidth guarantees, bounded delay) over pure packet-switched architectures remains an engineering challenge, especially for real-time audio and video.

Packet Delay, Loss, and End-to-End Throughput

  • Four Fundamental Sources of Packet Delay:

    • The total nodal delay experienced by a packet at a single router hop is modeled as:

dnodal=dproc+dqueue+dtrans+dpropd_{\text{nodal}} = d_{\text{proc}} + d_{\text{queue}} + d_{\text{trans}} + d_{\text{prop}}

  • Nodal Processing Delay (dprocd_{\text{proc}}): The time required to examine the packet header for bit-level errors and determine the appropriate outbound link using the forwarding table; typically microseconds or less (<1 ms< 1\,\text{ms}).

  • Queueing Delay (dqueued_{\text{queue}}): The time a packet waits in the output buffer for preceding packets to be transmitted; depends dynamically on the congestion level of the router.

  • Transmission Delay (dtransd_{\text{trans}}): The time required to push all of the packet's bits onto the physical link:

dtrans=LRd_{\text{trans}} = \frac{L}{R}

  • Propagation Delay (dpropd_{\text{prop}}): The time required for a physical bit to travel from the beginning of the transmission medium to the receiving router at the physical link length dd and medium propagation speed ss:

dprop=dsd_{\text{prop}} = \frac{d}{s}

  • The propagation speed ss in common guided media is approximately 2×108 m/s2 \times 10^8\,\text{m/s} (roughly two-thirds the speed of light in a vacuum).

  • Distinction: Transmission delay (L/RL/R) depends solely on packet length and link bandwidth, independent of distance; propagation delay (d/sd/s) depends solely on physical distance and medium speed, independent of packet length and transmission rate.

    • Caravan Analogy for Transmission versus Propagation:

  • Caravan Analogy Parameters: A highway represents a physical transmission link; a toll booth represents a packet switch; an individual car represents a single bit; a caravan of 1010 cars represents a packet containing 10 bits10\,\text{bits}.

  • Scenario 1 (Slow transmission, standard propagation):

    • Caravan size =10 cars= 10\,\text{cars}.

    • Toll booth service rate =12 seconds per car= 12\,\text{seconds per car} (time to transmit one bit).

    • Highway length =100 km= 100\,\text{km}; propagation speed =100 km/hr= 100\,\text{km/hr}.

    • Time to service and push all 1010 cars through toll booth 1 (transmission delay):

10×12 seconds=120 seconds=2 minutes10 \times 12\,\text{seconds} = 120\,\text{seconds} = 2\,\text{minutes}

  • Propagation time for the last car to travel from toll booth 1 to toll booth 2:

100 km100 km/hr=1 hour=60 minutes\frac{100\,\text{km}}{100\,\text{km/hr}} = 1\,\text{hour} = 60\,\text{minutes}

  • Total time elapsed before the entire caravan lines up in front of toll booth 2:

2 minutes+60 minutes=62 minutes2\,\text{minutes} + 60\,\text{minutes} = 62\,\text{minutes}

  • Scenario 2 (Slow transmission, fast propagation):

    • Increase car propagation speed to 1000 km/hr1000\,\text{km/hr}; increase toll booth service time to 1 minute per car1\,\text{minute per car}.

    • Propagation time across the 100 km100\,\text{km} highway is:

100 km1000 km/hr=0.1 hours=6 minutes\frac{100\,\text{km}}{1000\,\text{km/hr}} = 0.1\,\text{hours} = 6\,\text{minutes}

* The first car finishes service at toll booth 1 after 1 minute1\,\text{minute}, travels for 6 minutes6\,\text{minutes}, and reaches toll booth 2 at the 7-minute7\text{-minute} mark.
* At 7 minutes7\,\text{minutes}, toll booth 1 has only serviced 77 cars; three cars remain queued at toll booth 1.
* This illustrates that the leading bits of a packet can arrive at a downstream switch before the trailing bits have been transmitted by the upstream switch.
  • Queueing Delay and Traffic Intensity:

    • Let RR denote the link transmission rate (bps\text{bps}), LL denote packet length (bits\text{bits}), and aa denote the average packet arrival rate (packets/second\text{packets/second}).

    • Traffic Intensity is defined as the dimensionless ratio:

Traffic Intensity=L×aR\text{Traffic Intensity} = \frac{L \times a}{R}

  • Behavioral Dynamics:

    • If L×aR≈0\frac{L \times a}{R} \approx 0, packets arrive infrequently, and average queueing delay is negligible.

    • As L×aR→1\frac{L \times a}{R} \to 1, queues form, and average queueing delay increases asymptotically.

    • If L×aR>1\frac{L \times a}{R} > 1, the rate of incoming data exceeds the service capacity; queues expand without bound, and theoretical average delay approaches infinity.

Average queueing delay as a function of traffic intensity
  • End-to-End Route Tracing via Traceroute:

    • The diagnostic program Traceroute measures round-trip delay to every router along a path to a destination host.

    • Sends three probe packets with an explicit Time-to-Live (TTL) field targeting router ii. Intermediate router ii decrements the TTL to zero, discards the packet, and returns an ICMP TTL-expired datagram.

    • The sending host calculates the elapsed round-trip delay between probe transmission and reply receipt for each of the three probes.

    • Sample route trace (gaia.cs.umass.edu to www.eurecom.fr):

    • Hops 1–3: Local access routing inside the university campus network (cs-gw.cs.umass.edu, border1-rt, cht-vbns.gw.umass.edu) with delays between 1 ms1\,\text{ms} and 6 ms6\,\text{ms}.

    • Hops 4–7: Regional and backbone educational networks (vbns.net, abilene.ucaid.edu) exhibiting delays between 11 ms11\,\text{ms} and 22 ms22\,\text{ms}.

    • Hop 7 to 8: Traversal of a trans-oceanic undersea cable, where delay jumps from 22 ms22\,\text{ms} to over 104 ms104\,\text{ms}.

    • Hops 9–16: European academic and regional networks (geant.net, renater.fr) with delays spanning 102 ms102\,\text{ms} to 135 ms135\,\text{ms}.

    • Hops 17–18: Designated by asterisks (* * *), indicating routers that silently dropped probes or suppressed ICMP reply generation.

    • Hop 19: Final target reached (fantasia.eurecom.fr) with recorded probe delays of 132 ms132\,\text{ms}, 128 ms128\,\text{ms}, and 136 ms136\,\text{ms}.

  • Packet Loss Mechanics:

    • Routers allocate finite physical memory buffers to hold packets queued for an output link.

    • If an arriving packet encounters a completely filled buffer, the router drops the packet (referred to as packet loss).

    • Lost packets may be detected and retransmitted by the sending host via reliable transport protocols (such as TCP), retransmitted by intermediate links, or permanently lost (in connectionless protocols such as UDP).

  • Throughput in Computer Networks:

    • Throughput: The rate (measured in bits/second\text{bits/second}) at which bits are successfully transferred between a sender and a receiver.

    • Instantaneous Throughput: The bit rate measured at an exact moment in time.

    • Average Throughput: The total volume of delivered data bits divided by the total time duration required to deliver them.

    • Bottleneck Link: The specific transmission link on an end-to-end path that provides the lowest transmission capacity, thereby constraining total end-to-end throughput.

    • For a path with server upload rate RsR_s and client download rate RcR_c, the average throughput is:

Throughput=min⁡(Rs,Rc)\text{Throughput} = \min(R_s, R_c)

  • In a shared network where 1010 concurrent connections share a common backbone link of transmission capacity RR, the per-connection share is R/10R/10, yielding:

Throughput=min⁡(Rs,Rc,R10)\text{Throughput} = \min\left(R_s, R_c, \frac{R}{10}\right)

  • In real-world Internet access environments, the client access rate RcR_c or server access rate RsR_s typically forms the constraining bottleneck rather than the high-capacity multiplexed backbone core.

Protocol Layering and Service Models

  • The Architectural Rationale for Layering:

    • Provides a structured framework to modularize complex networking systems containing hosts, routers, varied physical links, software applications, and network protocols.

    • Promotes modular system evolution; an individual layer's internal mechanisms can be updated or replaced without altering the operational interfaces with adjacent layers (e.g., updating gate check-in rules at an airport does not disrupt aircraft takeoff and routing).

  • The Internet Protocol Stack (Five-Layer Model):

    • Application Layer: Provides communication services directly to user-facing applications. Protocols include HTTP (Web access), SMTP (email transmission), FTP (file delivery), and DNS (hostname resolution). The Protocol Data Unit (PDU) is the Message.

    • Transport Layer: Provides logical process-to-process data delivery between distributed applications running on distinct end hosts. Protocols include TCP (reliable data transfer, flow control, congestion control) and UDP (unreliable, connectionless best-effort delivery). The PDU is the Segment.

    • Network Layer: Transports packets across the distributed network core, routing them from source host to destination host. Consists of the Internet Protocol (IP) and routing algorithms. The PDU is the Datagram.

    • Link Layer: Manages frame transmission across an individual communication link connecting adjacent network devices. Protocols include Ethernet, IEEE 802.11 (WiFi), and Point-to-Point Protocol (PPP). The PDU is the Frame.

    • Physical Layer: Transmits raw binary bits over the physical media as electrical voltages, optical pulses, or radio signals.

  • The ISO/OSI Reference Model (Seven-Layer Model):

    • Formulated by the International Organization for Standardization (ISO); incorporates two layers omitted from the five-layer Internet architecture:

    • Presentation Layer: Interprets the semantics and format of exchanged data; handles application-level encryption, data compression, and machine-specific data formatting.

    • Session Layer: Coordinates session synchronization, checkpointing, and communication state recovery.

    • In the Internet architecture, if an application requires presentation or session services, they must be implemented directly inside the application-layer software.

  • Data Encapsulation and Decapsulation Mechanics:

    • Encapsulation at Sending Host:

    • The application layer generates a message, MM.

    • The transport layer receives MM and prepends a transport header (HtH_t), creating a segment: [Ht∣M][H_t \mid M].

    • The network layer receives the segment and prepends a network header (HnH_n), creating a datagram: [Hn∣Ht∣M][H_n \mid H_t \mid M].

    • The link layer receives the datagram and prepends a link header (HlH_l), creating a frame: [Hl∣Hn∣Ht∣M][H_l \mid H_n \mid H_t \mid M].

    • The physical layer serializes the frame bits into physical signals.

    • Processing in Intermediate Packet Switches:

    • Link-Layer Switches: Process only physical and link layers (Layers 1 and 2); unwrap and parse the link header to route frames locally without inspecting network headers.

    • Routers: Process physical, link, and network layers (Layers 1, 2, and 3); decapsulate up to the IP datagram header to read the destination address, update TTL fields, select the outbound link, re-encapsulate into a new link-layer frame, and forward.

    • Decapsulation at Receiving Host: The destination host strips headers sequentially (HlH_l, then HnH_n, then HtH_t), verifying integrity at each step, before delivering the original application message MM to the receiving application process.

Network Security: Attack Vectors and Countermeasures

  • Fundamental Internet Security Landscape:

    • The early Internet was designed around an academic model of mutual trust among known users sharing a transparent network.

    • Modern security engineering must retrofit defenses across all layers of the protocol stack to address external attacks.

  • Malware Taxonomy:

    • Viruses: Malicious executable programs that spread by attaching to files, documents, or applications; require active human intervention (such as launching an infected email attachment) to replicate.

    • Worms: Autonomous, self-replicating malicious programs that spread without user action by actively exploiting known system vulnerabilities across network-connected hosts.

    • Spyware: Covert malware installed on a host that monitors keystrokes, tracks visited URLs, gathers credentials, and uploads confidential profiles to collection servers.

    • Botnets: Networks of compromised end systems (termed "zombies") under central command-and-control, coordinated to send spam or launch distributed denial-of-service attacks.

  • Denial of Service (DoS) and Distributed Denial of Service (DDoS):

    • Objective: Render an online service, server, or link infrastructure inaccessible to legitimate users by consuming memory, compute capacity, or access bandwidth.

    • DDoS Execution Steps:

    1. The attacker identifies a specific target system (e.g., web server).

    2. The attacker exploits vulnerabilities across distributed Internet hosts to enlist them into a coordinated botnet.

    3. The compromised hosts concurrently flood the target with bogus packets, exhausting its processing capacity and link bandwidth.

  • Packet Sniffing:

    • Deployed on shared or broadcast transmission environments (e.g., unswitched Ethernet, broadcast wireless channels).

    • A network interface card running in promiscuous mode reads and records every packet that traverses the local physical channel, including cleartext application payloads and authentication passwords.

    • Software tools such as Wireshark operate as packet sniffers for traffic inspection and analysis.

  • IP Spoofing:

    • Attackers inject packets into the network containing fabricated, false source IP addresses.

    • Used to bypass IP-based authentication controls or redirect response traffic toward an innocent third-party host.

History and Evolution of Computer Networking

  • 1961–1972: Early Packet-Switching Principles and the ARPAnet:

    • 1961: Leonard Kleinrock develops queueing theory models demonstrating the effectiveness of packet switching over circuit switching for bursty data.

    • 1964: Paul Baran proposes packet switching for robust military networks at the RAND Corporation.

    • 1967: The Advanced Research Projects Agency (ARPA / DARPA) conceives the ARPAnet plan.

    • 1969: The first operational ARPAnet node is installed; the initial four-node packet-switching network links UCLA, Stanford Research Institute (SRI), UC Santa Barbara (UCSB), and the University of Utah.

    • 1970: Norman Abramson creates ALOHAnet in Hawaii, pioneering wireless satellite packet switching.

    • 1972: The ARPAnet is publicly demonstrated; the Network Control Protocol (NCP) is deployed as the first host-to-host protocol; Ray Tomlinson authors the first networked email program; the ARPAnet expands to 15 nodes.

Early four-node ARPAnet network topology sketch from 1969
  • 1972–1980: Internetworking and Heterogeneous Networks:

    • 1974: Vinton Cerf and Robert Kahn publish the foundational architecture for interconnecting networks (internetworking).

    • Cerf and Kahn's Internetworking Principles:

    • Minimalism and Autonomy: Interconnected networks retain their internal architecture without requiring changes to participate in global internetworking.

    • Best-Effort Service Model: The core network routes packets on a best-effort basis without delivery guarantees.

    • Stateless Routers: Intermediate routers do not maintain per-connection state information.

    • Decentralized Control: Global coordination operates without a single centralized point of failure.

    • 1976: Robert Metcalfe invents Ethernet at Xerox PARC.

    • Late 1970s: Proprietary networking architectures proliferate, including DECnet, IBM's SNA, and XNA; early work begins on fixed-length packet switching (the precursor to ATM); the ARPAnet grows to 200 nodes by 1979.

  • 1980–1990: Protocol Standardization and Network Proliferation:

    • 1982: Simple Mail Transfer Protocol (SMTP) defined for email exchange.

    • January 1, 1983: Official deployment of TCP/IP across ARPAnet, completely replacing NCP ("flag day").

    • 1983: Domain Name System (DNS) defined for mapping hostnames to IP addresses.

    • 1985: Formal specification of the File Transfer Protocol (FTP).

    • 1988: Van Jacobson develops TCP congestion control algorithms to stabilize network throughput.

    • New Networks: Expansion of major national networks, including CSNET, BITNET, NSFNET, and France's Minitel system.

    • By the end of the decade, over 100,000 hosts are connected across an expanding confederation of networks.

  • 1990–2000s: The Commercial World Wide Web, New Applications, and Core Growth:

    • Early 1990s: ARPAnet is officially decommissioned.

    • 1991: The National Science Foundation (NSF) lifts restrictions on commercial traffic over NSFNET (NSFNET is subsequently decommissioned in 1995).

    • Early 1990s: Invention of the World Wide Web, building on hypertext concepts formulated by Vannevar Bush (1945) and Ted Nelson (1960s); Tim Berners-Lee invents HTML, HTTP, the initial web server software, and the first web browser at CERN.

    • 1994: Marc Andreessen develops the Mosaic browser at NCSA, subsequently founding Netscape.

    • Late 1990s–2000s: Commercialization of the Web accelerates; emerging network applications include instant messaging and peer-to-peer (P2P) file sharing (e.g., Napster, Gnutella, BitTorrent); network security becomes a primary architectural concern; global host counts reach 50 million with over 100 million users; core backbone links scale to gigabit-per-second (Gbps\text{Gbps}) capacities.

  • 2005 to the Present: Pervasive High-Speed Wireless, Cloud Infrastructure, and Private Backbones:

    • Internet host populations surpass 750 million devices, driven by mobile smartphones and tablets.

    • Broadband access and high-speed wireless connectivity become ubiquitous.

    • Online social networks expand to massive scale (e.g., Facebook approaching one billion users).

    • Large service and content providers (e.g., Google, Microsoft) build private global networks, bypassing the public Internet transit core to connect their private data centers directly to edge networks for low-latency delivery.

    • Commercial enterprises, universities, and web applications migrate services into centralized cloud platforms (e.g., Amazon EC2).