TCP uses a windowing algorithm with a changing window size to constantly react to changing network conditions. The excellent analogy for multiplexing given in CNPS is air.
Networking Terms First, we need to get a few definitions out of the way. Think about an on-ramp to a freeway with the traffic lights on at rush hour, controlling the rate at which cars enter a crowded interstate. The source reacts to a packet loss by setting the threshold to half the current window, decreasing the current window to one packet, and entering the slow-start phase.
The window increases until a packet is lost, at which point the window size will sharply decrease. Where else can we look?
If a packet gets dropped, TCP shrinks the window size and decreases the amount of data allowed per transmission before the receiver has to acknowledge receipt of packets, effectively throttling traffic.
A better place for control would be at the gateway level where a global view is possible. Those require a good model and understanding of the particular type of traffic flow in a specific network. What happens when a buffer gets full?
When the calculated average QS AQS has grown above a certain level, the arriving packets are marked or dropped with a certain probability. Having a look to Random early detection gateways for congestion paper, tests show that the RED gateway is effective in controlling the average queue size.
As expected, the marked packets are more clustered with method 1 than with method 2. This is done by using a Exponential Weighted Moving Average which allows the average calculation to reflect only long-term several RTTs changes. But then as TCP ramps the window size up again, our buffer gets full, resulting in tail drop, and we start that whole cycle again.
One option for RED gateways is to measure the queue in bytes rather than in packets. VoIP requires packets to be delivered, and delivered on time. Identifying misbehaving users Because RED gateways randomly choose packets to be marked during congestion, RED gateways could easily identify which connections have received a significant fraction of the recently-marked packets.
Simulations show that network power ratio of throughput to delay is higher with RED gateways than with Drop Tail gateways. The average queue size should be kept low, while fluctuations in the actual size should be allowed to accommodate bursty traffic and transient congestion.
However, because RED gateways are designed to mark as few packets as possible, the overhead of setting the congestion indication bit is kept to a minimun. In either case, if desired, the RED gateway could be modified to give lower priority to those connections that receive a large fraction of the bandwidth during times of congestion.
If wq is set too low, the avg responds too slowly to changes in the actual queue size. The final packet-marking probability pb is calculated using an uniform random variable; this is achieved if the marking probability for each arriving packet is: The algorithm for computing the average queue size determines the degree of burstiness that will be allowed in the gateway queue.
These random numbers could be gotten from a table of random numbers stored in memory or could be computed fairly efficiently on a bit computer. Active Queue Management with non linear packet dropping function.
Some investigation has to be done with RED gateways that provide priority service for short control packets to reduce problems with compressed ACKs. TCP and IP work together to get packets of data from one place to another.
The RED gateway is designed for a network where a single marked or dropped packet is sufficient to signal the presence of congestion to the transport-layer protocol.
The weight wq should not be set too low, so that the calculated average queue length does not delay too long in reflecting increase in the actual queue length. As was saw, the list of packets marked by the RED gateway could be used to identify connections that are receiving a large fraction of the bandwidth; this information could be used to give such connections lower priority at the gateway.
However, RED gateways do not attempt to ensure that each connection receives the same fraction of the total throughput, and do not explicity control misbehaving users. A buffer is created at network interfaces to handle congestion. The Random Early Detection RED algorithm provides a way to randomly select packets to drop in order to prevent a full buffer and resulting tail-drop.
Then, when the average queue size was halfway between min. Because the gateway can monitor the size of the queue over time, the gateway is the appropriate agent to detect incipient congestion.
The rate at which RED gateways mark packets depend on the level of congestion. In this case, large packets are more likely to be dropped than smaller packets.
In basic queuing theory, we typically get around this by assuming a queue buffer in this case has infinite capacity. For controlling the average queue size in absence of cooperating sources, the RED gateways drop arriving packets when the average queue size exceeds some maximum threshold rather than setting a bit in the packet header.
The most effective place to detect congestion is in the gateway itself. With RED gateways the throughput for bursty traffic connections are close to the maximum possible throughput.
We can assume the process is stationaryor we can look into studying traffic the way we study fluid flow, typically using differential equations.Reference • “Random Early Detection Gateways for Congestion Avoidance” • Sally Floyd and Van Jacobson • Lawrence Berkeley Laboratory • University of California • August IEEE/ACM Transactions on Networking.
Random Early Detection Gateways for Congestion Avoidance. Jinyoung You CS, Network Architect. Contents.
Introduction of the Problem Previous works Design goals of the RED gateway The RED algorithm Simulation Results Calculation Parameter sensitivity Conclusions. Problem.
1 Computer Networking L-4 TCP 2 TCP Congestion Control • Congestion Control •RED • Assigned Reading • [FJ93] Random Early Detection Gateways for. Random Early Detection for Congestion Avoidance Sally Floyd Van Jacobson Drop tail Gateways End to End Congestion control may result in global synchronization.
Random Early Detection Gateways for Congestion Avoidance Sally Floyd and Van Jacobson IEEE/ACM Trans. Networking, Random Early Detection Gateways for Congestion Avoidance. Sally Floyd and Van Jacobson, IEEE Transactions on Networking, Vol.1, No. 4, (Aug ), pp Outline. Introduction Background: Definitions and Previous Work The RED Algorithm RED parameters RED simulation results.Download