Network layer protocols – Congestion problem & remedy

Network layers

The network layer is the third layer in OSI model. The network layer is needed because once the data packet go out from a LAN into an internet / networks, we then need an IP address for each data packet to travel and at this point MAC address is not enough and hence we need a network layer.

A transport layer protocol provides a logical communication between processes; however, A network layer protocols provides logical communication between hosts.

Network Layer provides the following:
1.    Unique address to every node which is known as IP address.
2.    Routing traffics between different nodes.

Physical and data link are connected to the next hop, whereas the network layer is end to end.

Congestion avoidance in network layer

Data link layer provides reliable communication from link to link; however, once the packet enters the internet cloud, it is very difficult to make sure that the data packet is not lost in the cloud. Therefore, the main concern for reliability is IP packet loss.

One of the main reasons for congestion and the loss of IP packets are due to the limited size of router buffers. All routers have two kinds of buffers, one at the input [inbuffer] and the other is at the output [outbuffer]. These buffers have certain limited size and if these buffers are completely occupied by data then there will be congestion in paths moving from this router and hence it will start dropping off IP packets. This will cause loss of IP packets due to congestion.

The solution to the problem of congestion and IP packet loss can be as follows:

1.    ARQ, Automatic repeat request
ARQ can be used in case an IP packet is not received at the other end of the host. For, example if an IP packet is not received at the other end of the host, that host will then send a request to the first host indicating that certain packet was not received and then the first host will retransmit that packet again. This cycle will continue until all packets are received at the other end. ARQ will be explained in details in upcoming lectures.

2.    Hypothetically, if the receiver host can somehow tell the sender host about the buffer configuration and path conditions, hence the sender can resize its packet.

Why there is a loss of Packet in IP cloud?

There can be to most important reason for the loss of packet in IP cloud. They are:

1. Packet loss happens due to the capacity of the buffers of routers. If the number of packets arriving at the router is greater than the capacity of the input buffer of the router, then some packets are going to be lost.
2. Packet may also be lost if TTL (Time to live) of an IP packet expires. When a router receives a packet, it decrements the value of TTL and forward it to the next hop. Eventually, this packet will be dropped by the router when TTL value will be equal to zero. This router will then send an acknowledgement back to the sender indicating that the packet was dropped due to zero TTL value. The sender will resend the packet with a higher value of TTL and this loop continues until the packet reaches the receiver and finally the receiver sends an acknowledgement that it has received the packet.

Problem: When the source sends a packet to the destination, the destination sends an acknowledgement packet back to the receiver. Suppose, the acknowledgement packet was dropped in a network then the sender will resend the packet, but this will increase the delay.

Remedy: The source / sender uses a timer to keep a track of a packet transmitted to a network. If the reply is not received within the specified time, then that packet will be sent again by the sender. The value of this time is estimated by the sender with the help of RTT (Round Trip Time) value.

Round trip time (RTT) is equal to the sum of the time a packet takes to reach the destination and the time an acknowledgement packet is received by the sender.

RTT value will vary from packet to packet, as packets always does not follow same path. Since we have different value of RTT, therefore the sender host needs to have a moving average of RTT’s and that value will be used by its timer.

Now suppose, the acknowledgement packet gets delayed, then the sender will re-transmit the packet again. This time there will be duplicate packets at the destination. The destination host will then drop one of the duplicate packets. Such kind of scenarios can also cause congestion in a network due to duplicate package.

A point to Remember: Data is originated by an application goes down to different layers of OSI model to physical layer and then into the cloud, reaches the destination and again goes from Physical layer to application and finally stored by it.

Click here to read more