4

I dont know if my research is enough or not but I could not find answers on the net.

My question is why not use bridges instead of routers. We know bridges can connect two different networks. Using bridges (or switches) we can transfer data from one network to another using the MAC address to identify the hosts.

Next, we say routers selects the best possible routes and send data to that line. It's just algorithm and so such algorithms like (Distance Vector Routing or Link State Routing) can be implemented in the bridges. Then bridges too can select best possible path and futhermore operates just using MAC addresses.

Now wouldn't this question the existence of router? And if routers don't exist then there would be no existence of IP address too.

Basically, my question is why need extra IP addressing to identify the devices. I also know people say IP is logical addressing and Bridges dont operate in Network Layer and such. So my question becomes why the exact need of Network Layer when Layer 2 can do those such things. What is the noticeable distinction between layer 2 and layer 3 such that layer 3 is essential?

I am well aware that I am missing some points. I just need to clarify my doubts with strong reasoning of need of IP addresses, Routers and Network Layer.

1
  • Did any answer help you? If so, you should accept the answer so that the question doesn't keep popping up forever, looking for an answer. Alternatively, you can post and accept your own answer.
    – Ron Maupin
    Jan 5, 2021 at 13:27

5 Answers 5

7

Imagine you are connected to a bridge with 3 ports. One port is connected to your host; one connects to networks to the west of you, and the last one connects to networks to the east of you.

Now suppose you want to send data to a device with a layer 2 address. How does your bridge know whether to forward the packet to the east or the west? Obviously, your bridge will look up the MAC address in its forwarding table and forward the packet out the proper interface.

Now imagine you are in a global network. Are you going to keep every MAC address in the world on your bridge? What happens as devices are turned on or off? If they move? How will you find unknown devices? Will you flood over the entire globe? You can't keep track of every device on a global network.

In short, layer 3 address provide routing information. Layer 3 address have a hierarchy that conveys location as well as identity. The network portion of an IP address tells you where the host is (topologically), and that information can be summarized, so your PC doesn't need to know how to forward data to every host on the Internet. That ability to scale routing makes a global Internet possible.

5

We know bridges can connect two different networks. Using bridges (or switches) we can transfer data from one network to another using the MAC address to identify the hosts.

That just isn't true. A bridge is a transparent, layer-2 device that works on a LAN. It knows nothing about other LANs or layer-3, e.g. IP, and there is no hierarchy. Every host on a LAN is a peer, and all broadcast traffic reaches, and interrupts, every host on a LAN, even if the traffic is not meant for that host, and LANs can grow large enough that they break because of all the broadcast traffic. (LANs are also know as broadcast domains.)

A router is required to get traffic from one LAN to another LAN because there is no concept of different LANs at layer-2. You need to have layer-3 to distinguish between LANs. The Internet routing table is huge, over 600,000 routes to different networks, but layer-2 would require individual addresses for all the hosts on each of those networks, and that just isn't practical.

There is very little security on a LAN, but it is very easy to add security between LANs at layer-3.

Layer-2 will get you from one host to a different host on the same LAN, but you need layer-3 to get from on LAN to another LAN, and you need layer-4 to get to the process within a host.

1

Something to realize is that the networking stack we all know and love today was NOT designed as a single coherent whole. The groups who designed IP and Ethernet were seperate groups designing separate standards to fulfill different goals.

Ethernet doesn't scale, there are a few reasons for this.

  1. Ethernet addresses do not reflect the host's network location. So a bridge must retain a full list of known network addresses.
  2. Ethernet has broadcasts (and "multicasts" that are effectively broadcast) and many protocol implementations running over Ethernet rely on them. As a network grows larger broadcasts become an ever-greater proportion of traffic.
  3. The data-plane learning technique used to locate hosts creates several problems. Firstly it requires a loop-free topology which doesn't scale. Secondly it means that some proportion of Unicast traffic can end up getting treated as broadcast traffic (which brings us back to point 2)

Why does it suffer from these limitations? because fundamentally it was never designed for building large networks. Could these problems be fixed with enough effort? absolutely but people don't because we already have IP.


IP was designed from the start to interconnect a range of disparate networks into one mega-network. As such it was designed from the start around explicitly assigned addresses that would reflect which network a host is on. It was designed from the start around allowing multiple paths. That is not to say there weren't/aren't growing pains, but IP was fundamentally aiming at large networks.

Yes having two source addresses and two destination addresses in every packet could be seen as wasteful, it is probably not how you would design a network if you were designing it from scratch as a coherent whole, but that is not how our current network stack came to be. It is two protocols designed by different groups for different reasons stacked on top of each other.

0

By default between Layer-3 networks you will not receive any ARP from another Network.

This is firstly security protection, So you should use a Specific VLAN number for your finger Print.

1
  • 1
    The OP wants to get rid of L3 altogether - you wouldn't even need ARP.
    – Zac67
    Jan 5, 2019 at 9:22
0

Using bridge devices and layer2 switches in networks is limited to forwarding frames in same networks with the reference to mac -address table which mapped mac-address to switch ports ..

But in scenario where traffic has to forward to another networks . Network layer is essential to find best path towards destination for selected best path it's check routing table in layer3 devices and forwards packets accordinglly . When forwarding packets between different networks ,destination ip address is considered . When forwarding frames on same networks destination mac address is considered

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.