What happens when a user in Seattle requests a website in Australia? How, and along what paths, does that request travel to the Australian server? What networks does it cross, and where does it change between them? And once the request reaches the server, what path does the response follow back to the user? To answer these questions and explore the physical geography of the Internet for his book “Tubes,” Andrew Blum traveled to a number of the Internet’s key past and present sites.
In general, the chapters follow a pattern: Blum chooses a particular feature of the Internet — the laying of cables, switching between networks or serving of data from storage centers — and visits a relevant location, meets with an expert and briefly explains how it works. He emphasizes how amazing and mind-blowing it is that the Internet is a set of physical locations and pathways rather than an amorphous cloud.
The broadest reach of the Internet is, of course, the wires. Every network owns or leases capacity on wires, ranging from those running into an end user’s computer to the thick cord connecting continents across oceans. Blum watches a team run a new wire under the sidewalk and meets up with another group as an intercontinental wire comes ashore. He generously describes the scenes, down to the laborers who “skip into the surf and untie [floats] from the cable” as it comes onto the beach. Every location that’s connected to the Internet has a wire hooked up to it, but the central paths tend to converge at locations dictated by geography or history where most of the other wires already are, since the whole point is to connect to other networks. The site where the first transatlantic telegraph cable came ashore remains popular, though prices for data transfer have dropped from the incredible early telegraph rate of $10/word with a 10-word minimum.
The convergence of these paths can be seen clearly on the maps of Internet traffic drawn by the company TeleGeography, which Blum visits. The maps, small versions of which are available on telegeography.com, are well worth looking at. It’s startling to behold the enormous number of wires necessary to connect all of the world’s Internet users and servers, and TeleGeography’s maps illuminate this vast and hidden web. These, and other images, would have been a welcome addition to “Tubes.”
As signals travel long distances along these wires, they gradually weaken, so they need to be regularly amplified. Blum mentions this problem in passing a couple of times, talking about the electricity needed “to power the undersea repeaters that amplified the light signals,” but doesn’t explain how amplification happens.
This omission exemplifies “Tubes”’ major weakness: Blum’s treatment of his subject tends to the light and superficial. Similarly, a reader might wonder what determines the physical route followed by a given request. Although route selection is undoubtedly complicated, the reader comes away from “Tubes” with no understanding of the process.
So data travels along these wires, and eventually it needs to cross from one company’s network to another. In the beginning of the Internet, this happened at a regional hub like the Metropolitan Area Exchange (MAE-EAST). All of the networks would plug into one machine, which would direct data among them. These machines couldn’t keep up with the rapid growth in traffic, so eventually the Palo Alto Internet Exchange (PAIX) was founded on a different model: Just let the networks plug into one another.
Internet exchanges now operate around the world, and networks interconnect through a process called peering, with network engineers from various companies negotiating connections. If they can’t come to an agreement, then their customers can’t communicate. While many of the companies are household names, the number of unfamiliar companies is remarkable: Equinix, Level 3 and NTT, for instance, operate significant parts of the Internet’s infrastructure.
At the end of the line (from a user’s perspective) lies a server, which hosts the data. The server processes requests and ships responses back out, where they travel along the same architecture, though not necessarily the same path as the request. Blum complains that his visit to a Google data center is “kabuki,” a “farce,” and his criticisms are well justified. The company doesn’t let him see anything, and security clearance for entry involves retina scanning. Every word spoken by his engineer guides is carefully monitored by hovering pr people, leaving one engineer “sullen — preferring to say nothing at all than to risk stepping outside the narrow box PR had inscribed for him.” Facebook is more forthcoming, but industrywide the operation of servers remains absurdly secretive, leaving him with the impression “that I, and by extension you, can’t be trusted to understand what goes on inside.”
By the end of “Tubes,” the reader will be familiar with some of the key features of the Internet’s infrastructure, but Blum’s blasé treatment of his subject leaves the reader with only cursory knowledge of how those features fit together to form the Internet.