Internet Architecture: The Global Network
Internet architecture describes how the global network of networks is physically built and logically organized — from submarine cables spanning ocean floors to the peering agreements between providers that determine how your data reaches its destination. Understanding internet architecture is essential for grasping how CDN infrastructure is strategically positioned to minimize latency, and why CDN providers invest heavily in peering and interconnection relationships.
How Did the Internet Evolve From ARPANET?
The internet's architecture evolved through distinct phases, each expanding the network's reach and capability:
- ARPANET (1969-1983): The internet's predecessor connected four university sites in October 1969. Funded by the U.S. Defense Advanced Research Projects Agency (DARPA), ARPANET pioneered packet switching — breaking data into small packets that could take different routes to their destination. By 1981, ARPANET connected 213 hosts, primarily at universities and research labs.
- NSFNet (1985-1995): The National Science Foundation built a backbone network connecting five supercomputer centers, initially at 56 Kbps (upgraded to 1.5 Mbps T1, then 45 Mbps T3). NSFNet's acceptable use policy eventually opened the network to commercial traffic in 1991, triggering explosive growth.
- Commercial Internet (1991-present): When the NSFNet backbone was decommissioned in 1995, commercial ISPs took over. The privatization created the tiered ISP structure we have today, with competitive backbone providers connected through Internet Exchange Points (IXPs).
- Broadband Era (2000-2010): DSL and cable replaced dial-up, increasing typical home connections from 56 Kbps to 10-100 Mbps. This bandwidth increase enabled video streaming, rich web applications, and the first generation of cloud services.
- Cloud and Edge Era (2010-present): Hyperscale data centers (AWS, Google, Azure), fiber-to-the-home, 4G/5G mobile, and CDN edge networks have created a multi-layered architecture where content is served from distributed locations rather than central data centers.
What Are Internet Exchange Points?
Internet Exchange Points (IXPs) are the physical locations where different networks meet and exchange traffic directly. They are critical infrastructure for CDN performance because they enable direct peering between CDN networks and ISPs, bypassing transit providers:
| Metric | Edge Processing | Origin Processing | Improvement |
|---|---|---|---|
| Latency (p50) | 8–15 ms | 120–300 ms | 15–20x faster |
| Cache Hit Ratio | 85–95% | N/A | Reduces origin load |
| TLS Handshake | 1–5 ms (nearby PoP) | 50–150 ms | 10–30x faster |
| Bandwidth Cost | $0.01–0.04/GB | $0.05–0.12/GB | 3–5x cheaper |
| Availability | 99.99% (multi-PoP) | 99.9% (single DC) | 10x fewer outages |
- DE-CIX Frankfurt: The world's largest IXP by peak traffic (exceeding 15 Tbps), connecting 1,000+ networks
- AMS-IX Amsterdam: One of the oldest and largest IXPs, handling 12+ Tbps of peak traffic
- LINX London: Europe's oldest IXP, connecting 900+ networks
- Equinix IX: Operates 40+ IXPs globally, co-located with Equinix data centers
CDN providers strategically place their edge servers (Points of Presence, or PoPs) at or near major IXPs. This placement enables the CDN to peer directly with hundreds of ISPs, ensuring that cached content reaches users with minimal network hops. Understanding CDN network topology reveals how these peering relationships shape content delivery performance.
What Are ISP Tiers and How Do They Connect?
The internet is organized into three tiers of Internet Service Providers:
- Tier 1: Global backbone providers (Lumen/CenturyLink, NTT, Telia, GTT, Cogent) that can reach every destination on the internet without purchasing transit from any other network. Tier 1 providers have settlement-free peering agreements with each other — neither pays the other for exchanging traffic.
- Tier 2: Regional providers that have some settlement-free peering but purchase transit from Tier 1 networks to reach parts of the internet they can't reach directly. Most large national ISPs are Tier 2.
- Tier 3: Local ISPs that provide "last mile" connectivity to end users. They purchase transit from Tier 2 (or sometimes Tier 1) providers and generally don't engage in settlement-free peering.
How Does Physical Infrastructure Shape the Internet?
The internet's physical layer comprises several key components:
- Submarine Cables: Over 550 active submarine cables carry 99% of intercontinental data traffic. These fiber-optic cables span ocean floors, with modern cables like MAREA (Microsoft/Facebook, Virginia to Bilbao, 2018) offering 200+ Tbps capacity. A cable break can disrupt connectivity for entire regions — hence the importance of route diversity.
- Terrestrial Fiber: Long-haul fiber networks connect data centers, IXPs, and cities. Dark fiber (unused, unlit fiber) is leased by CDN providers and cloud companies to build private backbone networks with guaranteed capacity.
- Data Centers: Modern internet data centers range from small edge facilities (Equinix IBX, 500-5000 sqft) to hyperscale campuses (Google, AWS, Facebook) spanning hundreds of thousands of square feet with dedicated power substations. CDN PoPs are typically deployed in carrier-neutral colocation data centers where multiple ISPs have presence.
- Last Mile: The connection between the local ISP and the end user — fiber-to-the-home (FTTH), cable (DOCSIS), DSL, fixed wireless, or cellular (4G/5G). Last-mile technology is often the primary bottleneck in end-user performance.
How Do CDNs Fit Into Internet Architecture?
CDN providers occupy a unique position in internet architecture — they are not ISPs in the traditional sense, but they operate one of the largest categories of networks on the internet:
- Massive Peering: Cloudflare peers at 330+ IXPs globally and has direct interconnects with 12,800+ networks. Akamai deploys servers inside 1,400+ networks worldwide. This extensive peering means CDN traffic rarely traverses paid transit links, reducing costs and latency.
- Embedded CDN: Some CDN providers (notably Akamai's Aura and Netflix's Open Connect) place cache servers directly inside ISP networks. A cable customer requesting a Netflix show gets content served from a cache appliance literally inside their ISP's facility — achieving single-digit millisecond latency.
- Private Backbone: Major CDN providers operate their own global backbone networks (Cloudflare's backbone spans 100+ cities) alongside their edge presence. This gives them control over inter-PoP routing, reducing dependency on the public internet for cache fill and content routing.