Part 3205 min total12 chapters

Networks

From the cable to NVLink (with Kumar Pratik).

From here on, a different voice takes the lead.

The earlier parts built the mental model: Compute, Store, Network as the three pillars; programs becoming processes; CPUs executing one instruction at a time; memory arranged in a hierarchy of speed.

This part follows one of those pillars all the way to the floor. It is written by Kumar Pratik, co-founder of GeekyAnts and a working engineer who has spent more than a decade building, breaking, and explaining networks. He writes longer than I do. He goes deeper than I do. He will not let you skip the homework.

How to Read These Ten Chapters

Read them slowly. Some are short. Most are not. Every one ends with exercises that are the actual point of the chapter, the writing is just the prompt.

When you finish, you will have followed a single packet from the metal contacts of an RJ45 jack, through glass on the seabed, up through OSI layers, into a kernel ring buffer, across a syscall boundary, and into the application that finally consumed it. You will also know why the most valuable company in the world is currently NVIDIA, and why that has very little to do with GPUs and very much to do with how chips talk to each other.

A Few Rules Before You Begin

Do not trust this book. Validate everything in it. If you find something that is wrong, that is good. Most learning happens during validation, not consumption.

Do not skip the struggle. The chapters end with questions and homework. The homework is not optional. It is the actual book. The chapters are just the prompts.

Do not ask other people for the answer. Ask them better questions. Pratik will explain why.

Turn the page.

Chapters in This Part

Chapter 1· 15 min

The Cable You Plugged In Without Thinking

Eight wires, two pairs that work, two LEDs that tell you everything. The hidden engineering inside an Ethernet cable.

Chapter 2· 15 min

Copper, Fiber, and the Limits of Distance

Why copper stops at 100 meters, how a glass thread carries light for 100 kilometers, and why your packet to America is touching the floor of an ocean.

Chapter 3· 14 min

The Hidden Journey of a Packet

60 ms is the physics floor. 200 ms is what you measured. Where do the missing 140 ms go?

Chapter 4· 16 min

OSI, TCP/IP, and Why We Compressed Seven Into Four

Two reference models, one journey. How a network request gets wrapped on the way down and unwrapped on the way up.

Chapter 5· 14 min

MAC Addresses, Hubs, and the Quiet Power of Switches

Layer 2: hardware addresses, learning switches, ARP, and the assumptions that quietly hold a network together.

Chapter 6· 25 min

IP, Subnets, DHCP, NAT: Building a Network That Works

Layer 3: the addresses that span the planet, the masks that slice them, and the hacks that kept IPv4 alive.

Chapter 7· 15 min

Ports: How Data Finds Your Application

Layer 4: how a 16-bit number on top of an IP address routes a packet to the right process on a busy machine.

Chapter 8· 20 min

From Wire to RAM: Buffers, Interrupts, and System Calls

The last mile of a packet: how a string in your application program is actually electrical pulses, an interrupt, a buffer, and a syscall.

Chapter 9· 20 min

Finding the Real Bottleneck

Your 1 Gbps link is running at 80 Mbps. The cable is fine. The wire is fine. The ISP is fine. Now what?

Chapter 10· 11 min

NVLink and the New Hardware Game

Why the most valuable company in the world isn't winning because GPUs exist — it's winning because of how fast its GPUs can talk to each other.

Chapter 11· 18 min

DNS, the Quiet Backbone

From the /etc/hosts file you've never opened to a global hierarchy of caches answering billions of lookups a second. The most important system you never think about.

Chapter 12· 22 min

Routing, BGP, and the Internet's Geography

Why two servers can share an IP address, who decides the path your packet takes, and why an undersea cable in the wrong place is a strategic asset.

Part 3: Networks | Junior2Senior.dev | Junior2Senior.dev