The RED algorithm is also known as Random Early Discard or Random Early Drop. This algorithm works at the router level. Congestion in computer networks must be discovered and fought from an early stage in order not to have an overloaded queue at the exit, which would eventually lead to blockage of the entire network.
The goals we must have if we are confronted with a congested network are:
1. An average tail level as small as possible
2. Lowest network delays
3. Value transmission speed
4. Getting a higher transfer rate
5. Enjoy the global synchronization that leads to the inefficient use of bandwidth
The congestion avoidance mechanism consists of ways of “predicting” the time at which congestion can occur and then using mechanisms to reduce the rate of packet transmission. And all this before you start removing (dropping) data packets.
The basic idea of the RED algorithm (Random Early Detection) is to announce the sources of congestion by marking and discarding packages before the congestion phenomenon occurs. For this, RED is based on controlling the average tail dimension, fitting it between two thresholds: min threshold and max threshold. The RED algorithm is designed for networks where the marking of a packet is sufficient to signal the presence of congestion. This control mechanism monitors the average size for each exit queue and chooses in a random way in which connection is to be notified of congestion.
The RED algorithm has some weaknesses such as:
- RED does not work well when the average quote size exceeds the max threshold resulting in a decrease in transfer rate and a high rate of packet throw.
- The average tail length varies according to the congestion level and can reach very close to the min threshold or the max threshold.
Our goal is to solve these problems with minimal changes to the RED algorithm. This proposal maintains the basic structure of the RED algorithm and only adjusts the maxp parameter to maintain the tail dimension between the minth and maxth thresholds. This proposal maintains the basic structure of the RED algorithm and only adjusts the maxp parameter to maintain the tail dimension between the minth and maxth thresholds. Thus, Adaptive RED reduces packet dropout and delays due to queuing queues. Early Random Detection (RED) is an effective mechanism for avoiding congestion at the gateway level, in conjunction with the network transport protocol.
The RED algorithm has the advantage of controlling congestion at the transport level and works with the current TCP/IP transport protocols.
The main purpose of the algorithm is to minimize packet loss and delays in routers, avoid global sourcing of sources, maintain a high degree of use of available power, and avoid the disadvantage of sudden transmission. RED is an effective mechanism for avoiding congestion at gateways in cooperation with transport protocols. It is a relatively simple algorithm that can be implemented in high-speed networks.
There are issues to solve in connection with this algorithm: determining the optimum average size of the queue to maximize the number of packets correctly transmitted and minimize delays for different network configurations, RED implementation with other TCP outbound transport protocols.