Internet Structure: Network Edge, Core, and Switching Mechanics, and Evolution

Overview of Network Edge and Core

  • The study of the Internet involves an in-depth look into its structure and functioning, specifically focusing on the Network Edge, the Network Core, and the mechanisms of Packet Switching.

    • Network core are the interconnected routers, the webs of the networks of networks for example the spectrum router connected to my nighthawk router.

  • This material covers foundational concepts corresponding to Chapter 1.2 through 1.3 of the course curriculum.

The Network Edge: Hosts and Access Networks

  • The network edge consists of end systems or hosts, which are categorized into clients and servers.

  • Clients include devices such as mobile phones, laptops, and smart home appliances.

  • Servers are often located in specialized facilities known as data centers.

  • Access networks refer to the physical media—wired or wireless communication links—that connect the network edge devices to the first router (also known as the edge router).

Host Perspective on Data Transmission

  • When a host sends data, it takes the application message and breaks it down into smaller units called packets.

  • Each packet has a length defined as L bitsL \text{ bits}.

  • The host transmits these packets into the access network at a transmission rate RR, which is also referred to as link capacity or bandwidth. The units for RR are bits per second (bps\text{bps}).

  • The time required to transmit a single packet into the link is calculated by the formula:

Time to transmit a packet=L (bits)R (bits/sec)\text{Time to transmit a packet} = \frac{L \text{ (bits)}}{R \text{ (bits/sec)}}

  • For example, if a packet length L=10 KbitsL = 10 \text{ Kbits} and the transmission rate R=100 MbpsR = 100 \text{ Mbps}, the one-hop transmission delay is 0.1 msec0.1 \text{ msec}.

Access Network Technologies

  • Cable-Based Access:

    • Uses Frequency Division Multiplexing (FDM), where different channels (video, data, control) are transmitted in different frequency bands.

      • All homes on a street are connected to one cable headend and these are kept from mingling because each home is on a different frequency to avoid collisions.

    • Utilizes coaxial cable and a cable headend to connect to the ISP.

      • The ISP is connected to a cable headend which is durable and capable of intaking multiple inputs, the cable is connected to a splitter which each home will have, this device allows for both internet and cable (for TVs) to intake both and divide them.

    • Transmission rates are asymmetric:

      • Downstream: 40 Mbps40 \text{ Mbps} to 1.2 Gbps1.2 \text{ Gbps}.

      • Upstream: 30 Mbps30 \text{ Mbps} to 100 Mbps100 \text{ Mbps}.

    • This is a shared access medium among multiple hosts.

  • Wireless Access Networks:

    • Wireless Local Area Networks (WLANs): Typically based on 802.11b/g/n802.11b/g/n (WiFi) standards. These operate within a local range (~100 ft100 \text{ ft}) with transmission rates ranging from 11 Mbps11 \text{ Mbps} to 450 Mbps450 \text{ Mbps}.

    • Wide-Area Cellular Networks: Provided by cellular operators (e.g., 3G and 4G). These have a range of tens of miles and provide transmission rates in the tens of Mbps. These operate as shared access networks.

  • Residential Access:

    • Residences often use a single device that combines multiple functions: cable modem, router, firewall, Network Address Translation (NAT), wired Ethernet (1 Gbps1 \text{ Gbps}), and a wireless access point (54 Mbps54 \text{ Mbps} or 450 Mbps450 \text{ Mbps}).

  • Enterprise Access:

    • Used by companies, universities, and government agencies.

    • Features a mix of wired (Ethernet) and wireless (WiFi) technologies.

    • Infrastructure consists of a hierarchy of switches and routers connecting to the ISP.

  • Datacenter Access:

    • Consists of hundreds to thousands of servers interconnected and connected to the Internet.

    • Utilizes high-bandwidth wired links ranging from 1 Gbps1 \text{ Gbps} to hundreds of Gbps.

    • Uses intricate topologies of switches and routers designed to maximize bandwidth and minimize latency.

The Network Core and Packet Switching

  • The network core is a mesh of interconnected routers that forms the backbone of the Internet.

  • Its primary function is to move packets from the source to the destination using two key functionalities:

    • Forwarding (Switching): A local action within a router where an arriving packet is moved from the router's input link to the appropriate output link. This is guided by a local forwarding table that maps header values (destination addresses) to specific output links.

    • Routing: A global action that determines the end-to-end path taken by packets from the source to the destination. This is handled by routing algorithms.

Mechanisms of Packet Switching

  • Store-and-Forward:

    • A router must receive the entire packet (all LL bits) before it can begin to transmit the first bit onto the next output link.

    • This introduces a transmission delay of L/RL/R seconds at each hop.

  • Queueing and Packet Loss:

    • Queueing occurs when the arrival rate of packets to a link exceeds the link's transmission rate for a period of time.

    • Packets wait in a buffer (memory) to be transmitted.

    • If the router's buffer fills up, arriving packets may be dropped, resulting in packet loss.

Comparison: Packet Switching vs. Circuit Switching

  • Circuit Switching:

    • Resources (bandwidth, switch buffers) are dedicated and reserved for the duration of a "call" between source and destination.

    • Used traditionally in telephone networks.

    • Guarantees performance but is inefficient because resources remain idle if not used by the call.

    • Sharing is not permitted.

  • Packet Switching:

    • Great for "bursty" data due to efficient resource sharing through statistical multiplexing.

    • Simpler to set up because no a priori reservations are required.

    • Congestion can occur, leading to delays or loss, requiring protocols for reliable data transfer and congestion control.

  • Numerical Example Comparison:

    • Scenario: A 1 Gbps1 \text{ Gbps} link where each user requires 100 Mbps100 \text{ Mbps} when active, but users are only active 10%10\% of the time.

    • Circuit Switching: Only 1010 users can be supported because each is allocated a fixed 100 Mbps100 \text{ Mbps}.

    • Packet Switching: With 3535 users, the probability that more than 1010 users are active at the same time is less than 0.00040.0004, allowing for much higher utilization.

Internet Structure: A Network of Networks

  • The Internet is a complex hierarchy of Internet Service Providers (ISPs).

  • Connecting every access ISP directly to every other ISP is not feasible as it would require O(N2)O(N^2) connections.

  • The Hierarchical Model:

    • Access ISPs: The periphery where hosts connect.

    • Regional ISPs: Connect access ISPs to larger, national/global ISPs.

    • Tier-1 ISPs: A small number of large, well-connected commercial networks (e.g., Sprint, AT&T, Level 3) providing national and international coverage.

    • Internet Exchange Points (IXPs): Physical locations where different ISPs peer with each other to exchange traffic.

    • Content Provider Networks: Private networks operated by companies like Google, Amazon, and Akamai. These networks connect their own data centers to the Internet, often bypassing Tier-1 and regional ISPs to reduce latency.

Examples of Infrastructure

  • Sprint (Tier-1 ISP): Utilizes Points of Presence (PoPs) to connect to customer networks and peering networks.

  • Google (Content Provider): Operates a global network involving over 100100 Points of Presence, extensive fiber networks, and Google Global Cache (GGC) edge nodes. They utilize major subsea cables such as FASTER (US, JP, TW), Unity (US, JP), Monet (US, BR), and SJC (JP, HK, SG).