The Network Core

1. The Network Core (First Image)

Mesh of Interconnected Routers: The network core consists of routers that
connect different parts of the internet. These routers form a complex,
interconnected structure that handles data transmission between different
networks.

Packet-Switching: This is a key concept in how the internet works. Instead

of sending a message as one large unit, it is broken down into smaller
chunks called packets. These packets travel independently from one router
to another until they reach their destination.

How it works: Each packet is forwarded from one router to the next
along a path, making its way from the source to the destination. The
transmission occurs at the full capacity of each link between routers.

2. Packet-Switching: Store-and-Forward (Second Image)

Store-and-Forward Transmission: This technique ensures that the entire
packet is received by a router before it is forwarded to the next one. In
other words, a router “stores” the packet before it forwards it to the next
destination. This adds a bit of delay.

Transmission Time Calculation:

L/R seconds is the time it takes to transmit a packet. Here:

L is the packet size in bits.0

R is the transmission rate in bits per second (bps).

For example, if L = 7.5 Mbits (megabits) and R = 1.5 Mbps (megabits per
second), it takes 5 seconds to send one packet over one hop (a single
step between routers).

End-to-End Delay: The total delay for a packet traveling from the source to
the destination is 2L/R (this assumes zero propagation delay, which is the
time it takes for a signal to travel through the physical medium).

3. Packet Switching: Queueing Delay and Loss (Third Image)

The Network Core 1

 Queueing Delay: Routers can only handle a certain amount of data at a
time, so packets often have to wait in a queue if the network is congested
or the link speed is slower.

Example: In the image, the link between two routers has a transmission
rate of 1.5 Mb/s (megabits per second), but the incoming data rate from
source A is 100 Mb/s. This creates a bottleneck, causing packets to
queue up.

Packet Loss: If the queue (buffer) becomes full, additional packets are
dropped and lost. This happens if the arrival rate of packets exceeds the
transmission rate for a significant period.

Why this matters: Dropped packets need to be retransmitted, causing

further delays and reducing the efficiency of the network.

4. Two Key Network-Core Functions: Routing and Forwarding

1. Routing:

Routing is the process that determines the path or route a packet (small
chunk of data) takes from the source to the destination.

This involves using routing algorithms, which calculate the best path to
the destination.

Think of it like a GPS deciding the best way to get from one city to
another.

2. Forwarding:

Once the route is determined, forwarding is the process of moving the

packet from one router's input to its appropriate output.

The router reads the destination address on the packet's header

(similar to reading a label on a package) and sends it out through the
correct link toward its destination.

5. Circuit Switching
1. Circuit Switching:

Circuit switching involves reserving end-to-end resources (like links or

circuits) for a "call" or communication between two devices.

The Network Core 2

 In the diagram, each link has multiple circuits, and a particular
communication uses a specific circuit (like a dedicated lane on a road).

No sharing: When the resources (circuits) are allocated to a call, no

other communication can use those circuits, even if they’re idle.

This type of switching guarantees a fixed and continuous connection

and is commonly used in traditional telephone networks.

6. FDM vs. TDM (Circuit Switching Types)

1. Frequency Division Multiplexing (FDM):

In FDM, different users are assigned different frequency bands. These

bands remain fixed for the entire duration of the communication.

It’s like each user having their own lane on a highway and driving at the
same time without interfering with each other.

2. Time Division Multiplexing (TDM):

In TDM, different users take turns using the same frequency, but they
are allocated distinct time slots.

Think of it like taking turns to use a single-lane road: each user has the
road for a specific amount of time before the next user gets their turn.

7. Packet Switching vs Circuit Switching

This slide compares two fundamental methods of transmitting data in a
network: packet switching and circuit switching.

1. Circuit Switching:

How it works: A dedicated communication path (or circuit) is established

between the sender and receiver for the entire communication session.
Analogy: Think of it like a phone call where the connection is exclusive between
two people, and no one else can use that line while the call is in progress.
Example on the slide:

You have a 1 Mbps link (1 megabit per second), and each user requires 100
kbps (kilobits per second) when active.
In circuit switching, only 10 users can use the network (since 1 Mbps = 1000
kbps, and 1000 kbps / 100 kbps = 10 users).

This is because each user gets a fixed, dedicated amount of bandwidth.

The Network Core 3

 2. Packet Switching:

How it works: Data is broken into smaller chunks called packets, and each
packet is sent independently through the network. The packets can take
different routes to reach their destination.
Analogy: It's like sending individual letters (packets) instead of making a phone
call. Each letter takes its own path, and the communication line isn't reserved
for just one user.

Example on the slide:

In packet switching, the link can support more than 10 users because users are
active only 10% of the time.
The slide says that with 35 users, the chance of having more than 10 users
active at the same time is very low (less than 0.0004 or 0.04%).

Key Point:
Packet switching allows more users to share the network than circuit switching,
making it more efficient in many cases.

Questions on the Slide:

How did we get the value 0.0004? This likely refers to a probability calculation
based on the activity levels of users.

What happens if there are more than 35 users? If there are more users, the
probability of multiple users being active at the same time increases, which
could lead to congestion or delays.

8. Is Packet Switching Better?

This slide digs deeper into the advantages and challenges of packet switching.

1. Benefits of Packet Switching:

Great for bursty data: Many applications (like web browsing or emails) don't
need constant data transmission. Packet switching is ideal for these "bursty"
data transfers, where data is sent in short bursts rather than a continuous
stream.
Resource sharing: Since packets are sent individually and not in a dedicated
line, more users can share the same network resources.

Simplicity: No need to set up a dedicated communication line before data can

be transmitted (unlike circuit switching).

The Network Core 4

 2. Drawbacks of Packet Switching:

Congestion can happen: If too many users are sending data at the same time,
packets can get delayed or even lost. This means that additional protocols
(rules) are needed to ensure data is transferred reliably, and congestion is
managed.

3. The Question: How to Provide Circuit-like Behavior?

For applications that need continuous data streams (like video calls or audio
streaming), we need to ensure a guaranteed amount of bandwidth. Achieving
this using packet switching is an ongoing challenge.

Bandwidth guarantees: These are necessary to avoid interruptions or delays in

services that rely on real-time data (like video conferencing).

9. Internet Structure - Network of Networks

This slide gives a brief overview of how the Internet is structured.

1. End Systems and ISPs:

End systems: These are the devices that people use to connect to the Internet,
such as your phone, laptop, or computer.

Access ISPs: These are companies or institutions (like residential ISPs,

university networks, or company networks) that provide Internet access to end
systems. Examples include Comcast, Verizon, or a university network.

2. ISPs Must Be Interconnected:

ISPs themselves need to be connected so that data can flow between any two
devices on the Internet, even if they are on different networks.

Example: If you're sending an email from your laptop in one country to

someone else's phone in another country, your message will travel through
different ISPs to reach the recipient.

3. Complexity of the Internet:

The Internet is not just one big network; it's a network of networks. This
structure has evolved due to various factors, including economics (costs of
infrastructure) and national policies (how different countries manage their
networks).

4. The Stepwise Approach:

The Network Core 5

 The slide ends by suggesting that the Internet's structure is complex and
requires a step-by-step explanation to fully understand how different networks
are connected and how they operate.

10. First Image (Option: Connect Each Access ISP to Every

Other Access ISP):
Problem Statement: It shows a hypothetical situation where all access ISPs
(Internet Service Providers) directly connect to every other ISP.

Key Insight: The image explains that connecting every ISP to every other ISP is
inefficient because it results in too many connections. The complexity of this
approach scales quadratically, denoted as O(N²), where N is the number of
access ISPs. This is impractical as the number of ISPs grows.

11. Second Image (Millions of Access ISPs):

Problem Statement: The next image poses a question: If there are millions of
access ISPs, how should they be connected? It visually depicts several access
networks (the small blue shapes labeled "access net") in a ring-like formation,
emphasizing that connecting all of them directly isn’t scalable.

Implication: This suggests that another, more efficient method is needed to

interconnect all these ISPs.

12. Third Image (Option: Connect Access ISPs to a Global ISP):

Solution: The third image offers a potential solution. Instead of each ISP
connecting directly to every other one, they connect to a global ISP, which acts
as a central hub or transit ISP.

Economic Agreement: This model operates based on economic agreements

between customer ISPs (access networks) and provider ISPs (global ISPs). The
global ISP helps facilitate the exchange of data between the access ISPs.

Key Insight: This approach simplifies the network architecture, reducing the
number of required connections by centralizing the traffic through a few global
ISPs, making the system more scalable and efficient.

13. First Image:

Internet Structure: Multiple Internet Service Providers (ISPs) exist, each
managing its own network, called an "access network." These ISPs (like ISP A
and ISP B in the image) serve customers by connecting them to the internet.

The Network Core 6

 Internet Exchange Point (IXP): The IXPs are special hubs where ISPs connect
and exchange data with each other. This allows users of one ISP to
communicate with users of another ISP.

Peering Link: This is the direct connection between ISPs through IXPs, allowing
data traffic to be exchanged.

14. Second Image:

The concept remains similar, showing three competing ISPs (A, B, and C). Each
has its own access network connecting users to the internet.

As these ISPs compete, they must still interconnect to enable communication

across networks, which happens via IXPs or direct connections between ISPs.

15. Third Image:

Regional Networks: A regional network is introduced here, which acts as an
intermediary layer that connects smaller access networks to ISPs. It makes it
easier for smaller networks to connect to the larger internet infrastructure.

The overall structure shows a more complex network where access networks
connect to ISPs, and ISPs are interconnected via IXPs, with regional networks
providing additional routing capabilities.

16. First Image (Network of Networks):

ISPs (Internet Service Providers): The internet is made up of various ISPs that
connect different networks. ISPs like ISP A and ISP B are responsible for
managing connections to end-users through access networks.
Content Provider Network: Companies such as Google, Microsoft, and Akamai
may run their own content delivery networks (CDNs). These networks are
designed to bring services and content closer to the end users for faster and
more reliable access.
IXP (Internet Exchange Point): An IXP is a physical location where different
networks (including ISPs and content providers) exchange internet traffic. This
helps improve performance and reduces costs.
Regional Networks and Access Networks: These networks help ISPs connect
to users in specific areas. Access networks are responsible for the last-mile
connection to the end-user, like your home internet connection.

17. Second Image (Tier-1 ISP: Sprint):

The Network Core 7

 Tier-1 ISP: This is a large ISP that has direct access to the global internet
backbone. Tier-1 ISPs, like Sprint, are well-connected to other Tier-1 ISPs and
don’t need to pay for transit, meaning they form the backbone of the internet.
POP (Point of Presence): A POP is a connection point where users and smaller
networks can access the backbone network of a Tier-1 ISP. It acts as a hub for
routing traffic to and from the main backbone and individual customers.

18. Third Image (Network Structure: ISPs, Google, and IXPs):

Tier-1 ISPs: These are the largest ISPs with global reach. Examples include
Level 3, Sprint, AT&T, and NTT. They form the core of the internet’s
infrastructure.
Google and other Content Providers: Some companies, like Google, have their
own private networks. These can directly connect to ISPs or IXPs, bypassing
some of the traditional internet pathways to offer faster and more direct
service.

Regional ISPs: These smaller ISPs connect to Tier-1 ISPs and serve more
localized areas, often connecting through IXPs.
Access ISPs: These are the ISPs that serve the end-users directly, providing
home or business internet service.
Key Points:
Tier-1 ISPs form the backbone of the internet, connecting with each other
globally.
Content providers (like Google) have their own networks to deliver content
closer to the users.

Regional ISPs and Access ISPs connect local users to the internet via larger
ISPs.
IXPs are important hubs where traffic is exchanged between different
networks.
This model ensures a scalable and robust global internet, where various players
collaborate to route data efficiently.

Summary
1. Access Network (Local Connection to ISP)

The Network Core 8

 Your access network is the part of your ISP that physically connects your
device to the internet. It’s your first hop when you send a request (like
browsing a website). Examples of access networks include:

DSL, Cable, Fiber (for home users),

Cellular networks (for mobile users).

Your traffic goes from your device (via Wi-Fi or Ethernet) to your
modem/router, which is connected to the ISP’s access network.

2. Regional Network (Sometimes, Not Always)

Not every ISP is large enough to be directly connected to the major hubs of the
internet. In many cases, smaller ISPs or those in rural areas connect to regional
networks.

Regional Networks: These are networks that cover a specific geographic

area and aggregate traffic from multiple small ISPs. They help bridge the
connection between small, local ISPs and the larger ISPs or the internet
backbone.

If your ISP is small, your request might travel through a regional network
before reaching larger infrastructure. If your ISP is large enough, it might skip
this step and go straight to an Internet Exchange Point (IXP) or connect with
other ISPs directly.

3. Your ISP (Core Network)

Once the traffic reaches your ISP’s core network, the ISP processes and routes
the traffic to the next destination. ISPs manage a lot of infrastructure to direct
traffic through their own network.

At this point, your ISP has several options:

If the destination is peered with them at an IXP or has a direct connection

(e.g., a large content provider), your ISP will route the traffic directly to the
IXP or CDN.

If the destination isn’t peered directly, your ISP will send the traffic to the
internet backbone or another ISP via transit agreements.

4. IXP (Connecting ISPs and Networks)

Once your traffic reaches an Internet Exchange Point (IXP), it could take
different paths depending on where your data needs to go:

The Network Core 9

 Peering with Content Providers: Many large companies like Google,
Facebook, Netflix, etc., have their own Content Delivery Networks (CDNs)
and edge servers located at IXPs or within the ISP’s infrastructure. If the
content you’re requesting (e.g., a YouTube video) is hosted by one of these
large content providers, your ISP can directly exchange traffic with them at
the IXP. In this case, your data doesn’t need to travel far since it can be
delivered from a local CDN server.

Peering with Other ISPs: If the content isn’t hosted by a large company or
CDN, your traffic may need to be routed through another ISP to reach its
destination. This happens if the destination is not peered with your ISP. At
the IXP, your ISP can exchange traffic with another ISP, and your request
will be forwarded through that ISP’s network to its final destination.

5. CDN (Content Delivery Network)

If you’re accessing content from a large content provider (e.g., Google, Netflix,
Facebook), there’s a good chance the data comes from a Content Delivery
Network (CDN), which is a specialized type of network that delivers content
from servers closer to your location:

CDN Servers are often located near IXPs or inside your ISP’s data centers,
reducing the distance that the data has to travel. This results in faster load
times and smoother experiences for things like video streaming or web
browsing.

The CDN delivers the requested content directly from these nearby servers,
bypassing the need to send your traffic across long distances.

6. Other ISPs (If No CDN or Direct Peer)

If the content you’re requesting isn’t on a CDN and doesn’t have a direct
peering arrangement with your ISP, your data may have to travel to another ISP.
This happens when:

Your ISP doesn’t have a direct connection to the content’s network.

The content is hosted on a smaller network or somewhere distant.

At the IXP, your ISP can hand off your traffic to another ISP that is directly
connected to the destination. This process repeats if needed, with the data
hopping between ISPs until it reaches the final server hosting the content.

Visualizing the Process:

The Network Core 10

 Scenario 1: Content from a Big Company (Using a CDN)
1. Your Device (Access Network) →

2. ISP Core Network →

3. IXP (Direct peering with a CDN) →

4. CDN server delivers the content directly to your ISP →

5. Your Device (Content delivered locally).

Scenario 2: Content from a Smaller Website (No CDN)

1. Your Device (Access Network) →

2. ISP Core Network →

3. IXP (Peering with another ISP) →

4. Other ISP (Takes over and routes the traffic further) →

5. Final Destination Server (Hosted by the smaller website’s server, maybe far
away).

Summary:
Access Network: Connects you (the end user) to your ISP.

Regional Network: May connect small ISPs to larger networks or directly to

IXPs.

ISP Core Network: Routes your traffic within the ISP’s infrastructure.

IXP: An exchange point where ISPs and networks meet to directly exchange
traffic. At this point, if you’re accessing a large content provider (like
Google), it might connect you directly to a nearby CDN server.

CDN: If content is hosted by a large provider, it could be delivered through a

nearby CDN, making the process faster.

Other ISPs: If the content isn’t directly connected at the IXP or available via
CDN, your ISP hands off the traffic to another ISP to reach the final
destination.

So, yes, in most cases, if you're accessing large, common content (like
YouTube or Netflix), your traffic goes from ISP → IXP → CDN. If it's not from a
large provider, it may go from ISP → IXP → another ISP, continuing the journey
across the global internet until it reaches the server you’re accessing.

The Network Core 11