Introduction to the OSI Model

I see… and the DNS server would obviously be a local DNS server w/ an A record replying w/ a local IP address on the same subnet.

It might be useful to see the OSI model in action. Here’s an example of telnet where the client connects to the server:

Wireshark capture: telnet client server

Here’s a packet from the client (192.168.12.1) to the server (192.168.12.2):

thank you Rene

Nice… Detailed Explanation.

Hi Rene,

The Ethernet Frame Come to Physical layer then encoded to bit and then bit to signal ??Also please do explain on PRESENTATION LAYER.I can’t understand your example .

br/
zaman

Hi Zaman,

That’s right. Everything from the data link layer is sent to the physical layer, gets encoded as a signal and is sent on the wire.

Everything that happens in the application, presentation and session layer is done by the application. That’s why for us as network engineers, these layers aren’t always very interesting.

The presentation layer has three roles:

  • Translation: when one person is speaking Chinese and the other English, we won't have much of a conversation. Computers use different character encoding sets like EBCDIC or ASCII. Translation is used to translate from one character encoding set to another.
  • Compression: the presentation layer can do some compression to reduce the data throughput.
  • Encryption: optional but possible, the presentation layer can do encryption. An example is SSL.

Hope this helps!

Rene

1 Like

Beautiful

Hi Rene,

Thanks for your great reply…

I can’t understand the line in a book for Transport layer " Performs error correction before retransmit"

and for Datalink layer “Performs error detection not correction”

Could you please help me on it .Thanks

br//
zaman

Hi Zaman,

Error correction before transmission doesn’t make much sense to me. TCP uses a sequence number and acknowledgment number, these can be used for retransmissions if something doesn’t make it to the other side.

The other thing that TCP does is calculating a checksum for the header. This allows the receiver to check if the TCP segment is OK or not.

Rene

Hi,

I was wondering what can be the issues, if we just dont use the MAC address.
What if all devices have only IP addresses. What do you think would happen then.

Here is what I feel:

  1. Every device would have to look up all the way till layer 3 to decide how to forward packets/frames, not just L2.
  2. Using MAC addresses to identify devices on a local network enables us to use any upper layer L3 protocol we want to use, if we used only IP,then we would be stuck with only IP or IPX or AppleTalk etc. We cant use any other way.

I am not satisfied by my thoughts !
Can you give some more suggestions ?

@Mansi,
Without MAC addresses you would have more problems to solve:

  1. Convenience would be a problem. DHCP would no longer be possible, because until you have an IP, you can’t communicate on a network. This means that every time you move a computer or device to a new network, you would have to configure it manually.

  2. Management would be a problem. Without setting up a global registry to ensure every IP address is unique from the factory (which would be impossible with IPv4 since there aren’t enough addresses), how would you ever be able to track down a mis-configured machine that has a conflicting IP address?

  3. Security would be a problem. How would you stop someone from plugging in a rogue machine to your network and spoofing an IP address without the possibility of layer-2 level security?

  4. Resiliency would be a problem. There are many services, such as GLBP, HSRP, and VRRP, that depend upon a separation of IP addresses from MAC addresses to provide highly available services.

I am sure there are plenty of other issues that I haven’t thought of yet :slight_smile:

Thanks Andrew.
However I have some queries about your first and second point.

  1. MAC addresses wont be anyways needed since DHCP server is going to broadcast the offers anyways.
  2. As far as gratuitous ARP is an issue, some other method can surely be developed to understand conflicts: eg send out a broadcast ie a L3 Broadcast…if the receiver sees the source IP as its own IP, then it will know of an IP conflict. So yes even this issue can be solved.

Thanks for the answers Andrew.

  1. How is a machine supposed to receive a broadcast? You are assuming it already has an IP address to get it. Even if a broadcast somehow worked without a transport protocol (it wouldn’t), you would have to have a DHCP server on every possible broadcast segment on a network. Most businesses don’t work that way–they use centralized DHCP servers with IP Helpers that forward DHCP requests.

Hi,

Very nice explanation. I have a basic query, similar to the above ones, but other way. What all problems if we dont have IP address at all, and we have only MAC addresses for communication?

Regards,
Durga Prasad

Durga,
The Internet would break :slight_smile:

Seriously, it would, and here’s why: Let’s say we tried to use MAC addresses as a layer 3 identity for communicating between devices at a distance (not on the same local network). MAC addresses are burned into the network cards by the manufacturer, and each card has a unique MAC (in theory). The problem comes when you are trying to keep track of who is where. The reason that the Internet is able to function now is primary due to a protocol called BGP that figures out the best way to get from one IP address to another.

Even though there are millions and millions of connected devices, BGP is able to function by using groups of addresses–most BGP providers will reject routes for any network smaller than a /24. Being able to group large numbers of IP addresses together and treat them the same is critical so the memory and processing power required to run BGP wouldn’t overwhelm a router’s resources. In the case of try to track individual addresses on a per machine level (like we are thinking about doing with MAC addresses), this would force BGP to deal with the equivalent of /32 routes. Since devices can easily move from one location to another, there just isn’t a way to “lump” groups of them together. There would just be too many objects to keep track of within BGP!

1 Like

Hi Andrew,

Thanks for clearing me.

Regards,
Durga Prasad

19 posts were merged into an existing topic: Introduction to the OSI Model

Hi,
Very valuable explanation . I would like to confirm one point . The first piece of information (The frame which is displayed before Ethernet II) has been added by Wireshark. I believe it is not part of OSI and here in this case OSI layers will be starting from Ethernet(II) as layer 1 (voltage signals ) we couldn’t capture for analyse . Please correct me if I am wrong.

Hi Sreenath,

That is correct, this information is added by Wireshark and not part of the OSI model. What we see in Wireshark is layer two (data link) and above.

Hi Rene,

Sonet \SDH belongs to which layer in OSI ? I believe it belongs to layer-1 , if so next concern is which layer 2 protocol it use for encapsulation like Ethernet interface use Ethernet frame (layer-2) frame to encapsulate , will it use sonet/sdh frame in layer-2 ? If possible please demonstrate a scenario about the packet flow of traffic from Ethernet source to Ethernet destination and if we have sonet\sdh networks in between it.

Thanks,
Sreejith