Classification of TCP Performance Schemes in Wireless Network

TCP (Transmission Control Protocol) is one of the main protocols of the Internet protocol suite. It lies between the Application and Network Layers which are used in providing reliable delivery services. It is a connection-oriented protocol for communications.

In this article, we will see performance schemes that are implemented to enhance the performance of TCP in wireless communication. So basic knowledge of TCP would be required to understand the functioning of the schemes.

TCP Enhancement for the Last Hop network can be classified into two categories:

Connection management-related approaches, which can further be divided into –
- Split Connection(I-TCP, METP etc)
- End-to-End Connection(Freeze TCP, TCPW, JTCP etc)
Wireless loss-related approaches, which can be further divided into-
- Localization(Snoop, SACK aware Snoop, SNACK-NS etc)
- Explicit Notification (ECN, ELN, EBSN etc)

Split Connection

Segments TCP into two parts at the Base Station- Mobile Host and Fixed Host

FH-MH=FH-BS+BS-MH
If more than one wireless links exist, then more than two TCP connection is needed

Working:

Indirect TCP is one of the first protocols that use this method.
The TCP connection is divided into three segments. The Mobile Host(MH) and Fixed Host(FH) are at the end with the Mobile Support Router (MSR) in the middle.
The MSR is a router located within a Base Station which acts as a messenger between the Fixed Host and Mobile Host. It carries packets from FH to MH.
This process helps in reducing congestion-related problems as it stops packets from propagating into the fixed network if such problems are present. It also comes into action during transmission errors or packet losses by similarly obstructing the packets from entering the fixed network.

Advantages:

BS-MH protocol can be optimized keeping wireless characteristics in mind.
Local recovery of wireless losses. Faster than normal TCP due to shorter RTT in a wireless link.
Allows MH to move to a new cell with little disruption

Disadvantages:

End to end Autonomy is violated.
Permanent data loss if BS crashes
Handoff latency is high due to state transfer
Not easily scalable as the buffer is needed for each TCP connection so BS may become overwhelmed. the smart buffering technique is required.
Extra copying of data occurs at the BS.
Increases end-to-end latency.
Cannot be used with IPSec.

End-to-End Connection

It is used to improve the performance of the system by fast retransmission of lost data packets in the same window without the cost of pipe exhaustion as is the case with Fast retransmitting and Fast Recovery.

Working:

It does not depend on the intermediate nodes. Modify the actual protocol keeping wireless characteristics in mind.
Sender and receiver attempt to determine the cause of packet loss.
- At sender: statistics based on RTT, window size, and loss pattern are received.
- At receiver: If determined by the receiver, it sends a notification to the sender.
Upon receipt of the notification or by self-evaluation, the sender takes appropriate action by reducing the congestion window if congestion is detected or else retransmitting the lost packet if the loss is due to error.

Improvements to TCP are based on this idea:

NACK (Negative Acknowledgement): NACK is a sender sider optimization which uses an extra field in the TCP header to signal the presence of a corrupt packet that was received and to perform the retransmission of this given packet at the earliest even if numerous packets in the same window were corrupt.
SACK (Selective Acknowledgement): In this process, the receiver first collects error-free data packets up to three such packets. After that, it starts transmitting those packets to the sender to be further processed.

Advantages:

It is highly scalable.
It is end-to-end semantics is preserved.
It can deal with congestion and wireless-related losses effectively.
It can be implemented without any help from the network.
It can work with IPSec.

Disadvantages:

Deployment of End-to-End connection is difficult.
Does not work well as yet. Traffic pattern changes frequently.

Localization

The localization approach is used to improve TCP performance in wireless environments employing the link layer transmission technique. It is a wireless-related approach for improving TCP performance.

Working:

This approach hides the wireless link from the sender. In this method, there is a local memory of wireless loss enabling quick recovery cause of a shorter propagation delay.
The localization approach uses Base Section (BS) to minimize the effect of the wireless error. It restricts TCP to respond mostly to congestion.
The BS stores unacknowledged packets. If an acknowledgement is received, then the packet is removed. If a packet loss is detected via duplicate acknowledgement, then
- If the packet is found, then discard the duplicate acknowledgement and retransmit
- If the packet is not found, then forward the duplicate acknowledgement and mark it as a congestion case
To avoid interference with TCP transmission
- Set the timer properly
- Limit the number of retransmissions at link layer
The localization approach is very useful when
- TCP retransmission times is high
- It provides in-order delivery

Advantages:

Deployment is easy in this process.
End-to-end autonomy is preserved.
Deals congestion and wireless losses effectively.

Disadvantages:

Low mobility.
Scalability is an issue caused by overwhelming of the BS

Explicit Notification

The explicit Notification Approach is a type of wireless loss-related approach to improving the TCP performance in wireless networks.

Working:

The approach is based on the theory that the intermediate nodes know better the cause of packet loss.
Now we know that ideal TCP behavior is,
- To reduce congestion window in response to congestion
- To simply retransmit lost packets when the loss is due to an error
The receiver sends the notification to the sender which is retrieved via routers. The sender then takes appropriate action depending on the nature of loss reported to notification.
There are many design options for this type of approach depending upon who sends the notification, how the notification is sent and how the notification is interpreted.

Advantages: