Ch.11: Routing in Detail

Inspired by: YouTube

In the previous post, we explored how a router helps two devices in different networks communicate. In this post, we will look at routing in much finer detail. We will trace exactly how a packet travels from a source device (like Laptop A) to a destination device (like Laptop X or an AWS server), observing the exact changes that happen to the Ethernet frame at each hop.

What Exactly is Routing?

When your frontend application makes an API request to a backend server, the data packet travels over the network. Since your local machine and the backend server (perhaps hosted on an AWS EC2 instance in a distant data center) are on entirely different networks, they cannot communicate directly.

Your machine does not know the exact physical location of the destination. Instead, it relies on a series of intermediate devices, like switches and routers, taking multiple "jumps" or "hops." This process of finding a path to the destination through multiple hops is called Routing.

Routing can occur in three main scenarios:

1. Within the same local network.
2. Between two different local networks managed by the same router.
3. Across the internet to an external network.

Before exploring these scenarios, let us understand the network setup we will be referencing.

The Network Setup

Wait, Where Does "Network 2" Exist in the Real World?

You might be wondering: "In my house, my devices connect to my router via Wi-Fi, and the router plugs directly into my Internet Service Provider. I do not have a 'Network 2' plugged into my router!"

You are completely right. The abstract setup with a "Network 2" is intentionally designed to demonstrate the theory of how a router passes data between two different local networks (Case 2) before routing it to the internet (Case 3).

In a standard home setup, you usually only have Network 1 (your local devices like laptops, phones, and smart TVs) and the Internet (WAN). However, a real-world "Network 2" managed by a single router exists in scenarios like:

• Guest Wi-Fi: When you enable a Guest Wi-Fi network on your home router, it creates an entirely separate subnet (Network 2) so guests cannot access your personal devices (Network 1). The router acts as the bridge between both!
• Corporate Offices: Large enterprise routers often manage multiple local networks simultaneously (e.g., HR on Network 1, Engineering on Network 2) to enforce security and isolation.

What is a Switch?

In past posts, we saw that a router operates at Layer 3 (Network Layer) and makes decisions based on IP addresses. A Switch, on the other hand, is a Layer 2 (Data Link Layer) device. It does not inspect IP addresses. Instead, it looks at the MAC address inside the Ethernet frame to make routing decisions.

When a device connects to a switch port, the switch records the MAC address associated with that specific port. Therefore, a switch can efficiently "unicast" (send directly) a frame to the correct port rather than broadcasting it to everyone, provided it knows the destination MAC address.

Case 1: Routing in the Same Network (Laptop A to Laptop D)

Let us assume Laptop A (10.0.0.3) wants to send a packet to Laptop D (10.0.0.5).

1. Subnet Check: A applies its subnet mask and determines that D is within the same network.
2. ARP Request: A does not know D's MAC address. It creates an Address Resolution Protocol (ARP) request: "What is the MAC address of IP 10.0.0.5?"
3. Broadcast: A sends this request to the Switch. Since A does not know the destination MAC, it broadcasts it. The Switch forwards the request to all connected ports (B, C, D, and the Router).
4. ARP Response: D receives the request and sees its own IP. D caches A's MAC address in its own ARP table to avoid future lookups. D then generates a response: "My MAC is D's MAC."
5. Unicast Delivery: D sends the response to the Switch, addressed specifically to A's MAC. The Switch, knowing A is on port pA, smartly forwards the response only to Laptop A.

Now that A has D's MAC address, it constructs the Ethernet frame:

Laptop A sends this frame to the Switch, which efficiently forwards it to port pD. The packet successfully reaches Laptop D.

Case 2: Routing to a Different Network (Laptop C to Laptop X)

Now, Laptop C (10.0.0.4) wants to send a packet to Laptop X (192.168.1.5).

1. Subnet Check: C applies its subnet mask and realizes X is on a completely different network.
2. Default Gateway: C must forward the packet to its Default Gateway (the Router at 10.0.0.1). If C does not have the Router's MAC address, it performs an ARP request just like in Case 1.
3. Frame to Router: C creates the initial frame and sends it.

4. Router Processing: The Router receives the frame, breaks it open, and sees its own MAC address as the destination. It then inspects the IP packet. The destination IP (192.168.1.5) is not its own. Applying its subnet masks, the Router determines the IP belongs to Network 2.
5. Router ARP: The Router now acts as a source in Network 2. If it does not know X's MAC, it sends an ARP request in Network 2 for 192.168.1.5.
6. Final Frame: Once it has X's MAC, the Router constructs a new frame and sends it.

Notice that the Source IP and Destination IP never change, but the MAC addresses are updated at the router hop.

Case 3: Routing Over the Internet (Laptop C to AWS EC2)

This is the most complex scenario. Laptop C (10.0.0.4) wants to reach an AWS EC2 instance (3.4.5.6).

1. Subnet Check: C realizes the destination is external and sends the frame to the Default Gateway (the Router), exactly as it did in Case 2.
2. Router Check: The Router receives the frame. It checks its routing tables and subnet masks and realizes the IP 3.4.5.6 belongs to neither Network 1 nor Network 2. It must forward the packet to the Internet.

Network Address Translation (NAT)

Here is where a critical transformation occurs. Laptop C's IP (10.0.0.4) is a Private IP. Private IPs are not routable on the public internet. If the router sent the packet with 10.0.0.4 as the source IP, the AWS server would not know how to reply, as there are millions of 10.0.0.4 devices globally.

To solve this, the Router performs NAT (Network Address Translation):

• It replaces the Private Source IP (10.0.0.4) with its own Public IP (2.3.4.5).
• It assigns a random port (e.g., 54362) and maps it to C's original source port (e.g., 3000) in its internal NAT table.

Network Edge vs. Network Core

Before the packet hops through "Router P" and "Router Q," it helps to know what those routers actually represent in the real world. The internet is conceptually split into two regions:

• Network Edge: Where end systems live, your laptop, phone, and the servers you talk to (like the AWS EC2 instance). These are the "hosts" that generate and consume data, sitting at the outer boundary of the network.
• Network Core: The mesh of interconnected routers that moves packets between edge devices. It is the "plumbing" of the internet, you never connect to it directly; you connect through it.

The Access Network: Your First Hop to the Internet

The access network is the link that physically connects an end system at the network edge (Laptop C) to the first router in the network core (the edge router, in our case the Home Router). In the real world, this link is provided by an ISP and takes different forms:

• Home: DSL, Cable, or Fiber (FTTH).
• Mobile: Cellular data (4G/5G).
• Enterprise: Ethernet or Wi-Fi within an office building.

How ISPs Talk to Each Other: IXPs

Your home router hands the packet to "Router P" (your ISP), but no single ISP owns the entire internet. Eventually, your ISP needs to hand the packet off to a different ISP closer to the destination. This handoff typically happens at an Internet Exchange Point (IXP): a physical facility where many ISPs plug their routers into a shared switch so they can exchange traffic directly, instead of paying a third party to carry it for them.

Think of an IXP as a meeting point: ISP A's router and ISP B's router sit in the same building/rack, so a packet can hop directly from A to B instead of taking a long detour through some other network. Large cities host massive IXPs (like DE-CIX in Frankfurt or NIXI in Mumbai) that connect hundreds of networks simultaneously.

BGP: How Routers Decide Which ISP to Use

Once traffic needs to cross from one ISP's network into another's, the ISPs need to agree on the best path between them. This is handled by BGP (Border Gateway Protocol), the protocol ISPs use to advertise: "I can reach these IP ranges, route traffic for them through me." When Router P finishes processing your packet, it consults BGP-learned routes to decide which neighboring network offers the best path toward 3.4.5.6, then forwards the packet accordingly. Every hop you'll see below (Router P to Router Q) is, in effect, one BGP-advertised path being followed.

Circuit Switching vs. Packet Switching

One more foundational idea before we continue: how does the network actually move data? There are two classic approaches.

• Circuit Switching: The network reserves a dedicated, exclusive path between source and destination for the entire session before any data is sent, like a traditional phone call. Once connected, that channel is reserved whether you're talking or sitting in silence. Resources are guaranteed but wasted when idle.
• Packet Switching: What the internet actually uses. There is no dedicated path; data is broken into packets, and each one is sent independently, sharing links with everyone else's traffic. Routers decide the next hop for each packet individually, which is why the response to our packet could take a completely different path back.

Packet switching is more efficient because links are only used when there is actual data to carry, no bandwidth is wasted holding a line open during silence. This is exactly why the internet can support billions of devices sharing the same infrastructure.

The Journey Across the Internet

Now, the packet begins its journey across multiple ISP routers (Router P and Router Q). At every single hop, the Source and Destination MAC addresses change, but the Destination IP remains constant.

• Hop 1 (Home Router to Router P):
  • Source MAC: Home Router's Public MAC
  • Dest MAC: Router P's MAC
• Hop 2 (Router P to Router Q):
  • Source MAC: Router P's MAC
  • Dest MAC: Router Q's MAC
• Hop 3 (Router Q to AWS Router):
  • Source MAC: Router Q's MAC
  • Dest MAC: AWS Router's MAC

The frame arriving at the AWS Router looks like this:

Final Hop to EC2

The AWS Router receives the packet, sees that 3.4.5.6 belongs to its own internal network, and performs a final ARP request to get the EC2 instance's MAC address. It updates the frame one last time:

The Response Path

When the EC2 instance replies, it sends the response to the destination IP 2.3.4.5 on port 54362. The packet hops back across the internet until it reaches your home Router. The Router checks its NAT table, translates 2.3.4.5:54362 back to the private IP 10.0.0.4:3000, replaces the destination MAC with Laptop C's MAC, and forwards it through the Switch to Laptop C.

Key Takeaway

Whenever a packet travels over a network, the Destination IP never changes. However, the MAC addresses (both Source and Destination) are replaced at every single hop. The Source IP may also change if the packet traverses a NAT-enabled router to enter the public internet.

Understanding these transformations is the foundation of mastering network routing!