China
Africa
Explore chapters and articles related to this topic
Quality of Service in Switch/Routers
Published in James Aweya, Designing Switch/Routers, 2023
RED [FLOYACM1993] is a congestion avoidance scheme that tries to prevent the oscillations between link under-utilization and over-utilization that occur during TCP global synchronization caused by tail drop. RED tries to control the queue length by randomly dropping TCP packets as buffers fill beyond a configured threshold level. As the buffer fills and packets are randomly dropped, the affected TCP sources will throttle their rates and comparatively a smaller number of the affected TCP sources will decrease their transmission rates at the same time (avoiding global synchronization). If the buffer continues to fill, a higher and higher percentage of packets is randomly dropped. Generally, RED greatly reduces the number of buffer overflows, tries to distribute packet loss fairly among the TCP sessions, and minimizes queue length while still absorbing traffic spikes.
Traffic Control
Published in Naoaki Yamanaka, High-Performance Backbone Network Technology, 2020
RED monitors the average queue length and starts to drop arriving packets on a probabilistic basis when the average queue length exceeds a certain threshold. To make a router start to drop packets before its buffer becomes full, RED gives an early congestion indication to each flow and avoids congestion collapse. However, RED does not provide fairness among competing flows. FRED brings fairness of a bandwidth allocation to RED by maintaining a per-flow state. FRED drops packets from flows that have had many packets dropped in the past and from flows that have longer queues than the average queue length.
QoS and MPLS
Published in Nam-Kee Tan, MPLS for Metropolitan Area Networks, 2004
One method of avoiding all these undesirable behaviors is to use a technique known as random early detection (RED). RED is a congestion avoidance mechanism that randomly drops packets even before a queue is full. RED drops packets with increasing probability, and as a result, TCP sessions slow down to the approximate rate of the output port bandwidth and the necessary average queue size is reduced.
The impact of global on-line information provision on transport networks and how random early detection can help
Published in Transportmetrica B: Transport Dynamics, 2019
H. Grzybowska, S. Willmott, S. T. Waller
The Transmission Control Protocol (TCP) (Cerf and Icahn 2005), one of the main protocols of IP networks, includes a transportation mechanism for delivering data between computers or between computers and terminals. Internet networks frequently get congested to the point of information loss in extreme cases (i.e. drop of data packets). In order to improve network efficiency and diminish the negative impact of congestion a RED algorithm can be applied. RED algorithm tries to prevent congestion rather than just reacting to it by dropping packets of data before the storage capacity at gateways is reached. It also includes randomisation to ensure that if data packets are dropped, they are dropped at the same loss rate at all connections. The literature on RED algorithm in TCP is vast. For examples on applications of RED in communication networks we refer the interested reader to Cerf and Icahn (2005), Lin and Morris (1997), Trinh and Molnar (2004), Floyd and Jacobson (1993), Sharma, Virtamo, and Lassila (2002). For examples on applications of RED in distributed computing systems with multiple schedulers we refer the interested reader to Brugnoli et al. (2007), Brugnoli, Heymann, and Senar (2010).
A congestion-aware decision-driven architecture for information-centric Internet-of-Things applications
Published in International Journal of Computers and Applications, 2022
The proposed decision-execution supporting architecture starts by employing a congestion-aware admission module. The presented admission module aimed at regulating the admission of the submitted decisions, thus, maintains a balanced load (i.e. balanced submitted decision requests) even during the high demanding environmental periods. For that purpose, this work integrates a priority-driven variant of the Random Early Detection (RED) algorithm to be exploited at the proposed admission module. Basically, RED algorithm is a well-known queuing discipline that is widely used in the networking field [18]. The RED algorithm basic idea is concentrated around continuous monitoring for the average queue size. Classical RED sets a unified drop rate using the average queue length calculated using a low pass filter. Packets will be dropped probabilistically depending on the thresholds’ settings of the queue. The classical RED variant sets two thresholds for the queue namely: minimum threshold and maximum threshold. RED starts marking/dropping the packets when the average queue size exceeds the pre-set thresholds. Packets are marked or dropped probabilistically when the average queue size is within the minimum and maximum thresholds. However, if the average queue size exceeds the maximum threshold then all packets are marked or dropped. In RED the packet marking/dropping probability is expressed as [19]: It could be inferred from (1) that, as the average queue size Qavg varies from minTH to maxTH, the packet marking/dropping probability Pd varies linearly from 0 to maxp. Several versions of the RED algorithm have been proposed in the literature, for example, Gentle Random Early Detection (GRED) [20], Fuzzy Gentle Random Early Detection (Fuzzy GRED) [21], Stabilized RED (SRED) [22].