BGP and Internet Architecture

1. The Evolution of the Internet: From Research to Global Commerce

The Internet didn't just appear out of nowhere. It started as a small, government-funded experiment and slowly but surely evolved into the massive, commercial system we all rely on today. Let's take a quick trip back in time to see how this happened.

A Brief Timeline of the Early Internet

• 1969: ARPANET is launched with just four nodes.
• 1972: The first public demonstration of ARPANET happens, and email is introduced. Yes, email is that old!
• 1983: The TCP/IP protocol suite becomes the standard for ARPANET. This was a huge step toward the modern Internet.
• 1986: NSFNET is launched to connect supercomputing centers, and it quickly becomes the new backbone.
• 1990: ARPANET is officially turned off, with its role completely taken over by NSFNET and other networks.
• 1995: The NSFNET backbone is decommissioned, marking the full transition to a privatized, commercial Internet.

ARPANET (Advanced Research Projects Agency Network)

Back in 1969, the U.S. Department of Defense kicked things off by launching ARPANET. It started with just four nodes, connecting places like the University of California, Los Angeles (UCLA), the Stanford Research Institute (SRI), the University of California, Santa Barbara (UCSB), and the University of Utah. These nodes were linked by circuits that were a blazing 56 kbps. (Can you imagine trying to stream Netflix on that today?!) Its main goal was pretty straightforward: to let researchers share expensive and scarce computing resources across different locations. It was never intended for commercial use. As its value became clear, it expanded like crazy to connect tons of universities and government agencies. However, its centralized design and limited capacity weren't built for explosive growth, leading to a lot of congestion and scalability headaches.

NSFNET (National Science Foundation Network)

By the mid-1980s, as ARPANET was buckling under the strain, the National Science Foundation (NSF) stepped in to build NSFNET. Its first job was to give researchers access to the nation's five supercomputing centers. But thanks to its much better architecture and speed, NSFNET quickly grew to become the main backbone of the growing Internet in the United States, effectively replacing ARPANET.

Its most important contribution was its three-tiered, hierarchical network architecture. This robust and scalable model is a big reason why the Internet could grow the way it did.

• Backbone: This was the core of the network, a high-speed backbone connecting major supercomputer centers. It saw some serious performance upgrades, jumping from 56 kbps to T1 (1.544 Mbps) and then T3 (45 Mbps) by 1991. These upgrades gave the network the massive data capacity it needed.
• Regional Networks: This was the crucial middle tier. These regional networks were set up to be the scalable bridge between individual campus networks (the end-institutions) and the high-speed backbone. This model was a game-changer because it solved the problem of connecting every single university directly to the core network. It enabled broad participation and super-fast expansion.
• End Institutions: This was the final tier, the ultimate consumers and producers of network traffic. This included universities, research labs, and government agencies like NASA. Researchers used E-mail service and protocols like FTP for sharing files and Telnet for remote access to supercomputers.

The Rise of the Commercial Internet (1995 onwards)

By the early 1990s, everyone could see the Internet's massive potential beyond just academia. This eventually led to the development of commercial Internet Service Providers (ISPs). A pivotal moment came in 1995 with the decommissioning of the NSFNET backbone. Don't think of this as a shutdown; it was a strategic handover to a new market of commercial Network Service Providers (NSPs) like MCI, Sprint, and AT&T. This act effectively privatized the core of the U.S. Internet, paving the way for its crazy growth.

With a bunch of competing commercial providers, a new challenge popped up: How do you make sure all these separate networks can talk to each other? The solution, initiated by the National Science Foundation, was the Network Access Point (NAP). These were strategically placed data centers where different ISPs could physically connect to exchange data—a practice called "peering."

Core Architectural Principles

So, what did we learn from this journey? The Internet as we know it today is built on a few key design principles that came from this evolution.

• Hierarchical Design: Remember the NSFNET's tiered system? That model is everywhere today. We have global backbones, regional networks, and local providers. This hierarchy is what allows for efficient and scalable Internet.
• Distributed in Nature: No single entity owns the Internet. It's a collection of tens of thousands of independent networks, called Autonomous Systems (ASes), that voluntarily connect to each other. This distributed model makes the Internet resilient.
• The Critical Role of the Backbone: Just like NSFNET had its super-fast backbone, every modern network needs a solid core to handle massive traffic loads. Think of a backbone as the main highway for data in a network.
• Public Interconnection Points (NAPs/IXPs): We saw how NAPs were created to solve the problem of public interconnection. This concept lives on today in the form of Internet Exchange Points (IXPs). They are the key to the Internet's decentralized, interconnected design.

2. The Architecture of the Modern Internet

The modern Internet works on a global scale because it's built on principles of hierarchy and distribution.

Understanding ISP Tiers Through a Delivery Company Analogy

Imagine you're sending a package to a friend across the world. You drop it off at a small local delivery service. That local service doesn't have their own infrastructure to cross oceans, so they hand your package to a bigger regional company. The regional company, in turn, relies on a global giant like FedEx to carry the package across continents. Finally, another local service picks it up from FedEx and delivers it to your friend's door.

The Internet's Delivery System: ISP Tiers

• Tier-1 ISPs: These are the global giants, the "FedEx" of the Internet. They own and operate massive, global networks and can reach any other network on the public Internet without paying anyone for access. They do this through "settlement-free peering."
• Tier-2 ISPs: Think of these as the regional delivery companies. They run significant national or regional networks but don't have a full global footprint. So, to connect their customers to the entire world, a Tier-2 ISP has to buy IP transit from one or more Tier-1 providers.
• Tier-3 ISPs: These are the local delivery services. They mostly operate at the local level, focusing on providing "last-mile" Internet access directly to you, the end-user.

How It All Works Together

Picture this: You click on a video hosted on a server in Europe. Here's how the Internet's "delivery system" gets it to you:

1. Your Tier-3 ISP (the local service) picks up your request from your device.
2. It hands the request to a Tier-2 ISP (the regional company) that covers your area.
3. The Tier-2 ISP sends it through a Tier-1 ISP (the FedEx), which carries it across the ocean via its massive global network.
4. On the other side, another Tier-2 or Tier-3 ISP delivers it to the server.
5. The video then travels all the way back to you through the same system!

3. Internet Relationships and Link Types

The connections between networks on the Internet aren't just random cables; they are governed by business agreements that determine how traffic flows and who pays.

• Customer-Provider: This is the most common type of connection. In this model, one network (the customer) pays another, typically larger, network (the provider) for access to the global Internet. This service is called IP Transit.
• Peering (Friends): Peering is a reciprocal relationship between two networks, usually of similar size. Under a peering agreement, both networks agree to exchange traffic directly between their respective sets of customers for free. It's important to remember that a peering link only gives you access to the routes within your peer's own network.

Quick Summary:

4. The Critical Role of Peering and Internet Exchange Points (IXPs)

Peering is absolutely essential for an efficient Internet. You can do it privately or, more commonly, at public Internet Exchange Points (IXPs). An IXP is a physical data center where hundreds of networks meet to connect. Instead of setting up tons of costly private links, a network operator can just set up a single connection to an IXP and then set up peering sessions with hundreds of other members (like Netflix, Google, Meta, and other ISPs) on the same shared infrastructure.

5. Submarine Cables: The Physical Backbone of the Global Internet

We often think of the Internet as being "in the cloud" or wireless, but that's not the full picture at all. The global Internet is actually at the bottom of the ocean! The honest truth is that over 95% of all international data—your video calls, streaming shows, and messages—travels through a massive network of physical, fiber-optic cables laid on the seafloor. These submarine cables are the true highways of the Internet, connecting continents.

Laying a cable across an ocean costs billions of dollars, so it's too big a job for just one company. They are built by consortia—groups of giant companies that team up, including big tech players and telecom leaders.