China
Africa
Explore chapters and articles related to this topic
Quality of Service in Switch/Routers
Published in James Aweya, Designing Switch/Routers, 2023
In order to mitigate the packet loss behavior of TCP, all TCP implementations use the TCP slow start mechanism (or its variant). Studies have found that if TCP starts transmitting its window of data relatively slowly and gradually increases its transmission rate to the full rate of the network path to the receiver, it can achieve a balanced transmission rate that does not overmatch the available network path resources. A TCP sender accomplishes this by doubling its transmission rate with each acknowledgment from the receiver and halving the transmission rate whenever there is data retransmission due to packet loss or receiver window exhaustion. Reference [RFC5681] describes the various TCP congestion control algorithms (slow start, congestion avoidance, fast retransmit, and fast recovery) used in practical TCP implementations.
Network Models
Published in Sunit Kumar Sen, Fieldbus and Networking in Process Automation, 2017
TCP uses a very helpful congestion control mechanism that can overcome any possible overwhelming of the receiver by the sender. This is a possibility for slow WAN links. TCP congestion control algorithms can adapt the sender to the network capacity at any point of time and thus avoid any potential congestion situation. TCP follows a variety of congestion control algorithms to avoid congestion: slow start, congestion avoidance, fast retransmit, and fast recovery.
Stability analysis of PD AQM control for delay models of TCP networks
Published in International Journal of Control, 2022
Adrian Puerto-Piña, Daniel Melchor-Aguilar
During the last decades, a lot of efforts have been focused to the analysis and design of Active Queue Management (AQM) schemes for supporting end-to-end Transmission Control Protocol (TCP) congestion control in networks. The fluid-flow delay model introduced in Hollot et al. (2002) for describing the behaviour of TCP/AQM networks has become a reference for investigating the qualitative properties of TCP/AQM networks and developing control theoretic design and analysis for the AQM. Thus, based on such a model, Proportional (P) and Proportional-Integral (PI) controllers were proposed in Hollot et al. (2002); Proportional Derivative (PD) control are introduced in Sun et al. (2003), Kim (2006) and Azadegan et al. (2015); controllers in Quet and Özbay (2004); while Yan et al. (2005) consider a variable structure control as AQM scheme. Due to their simplicity and easy implementation, several works have been devoted to compare and modify the P, PI and PD controllers for improving stability, robustness, and performance properties, see the survey papers Ryu et al. (2003) and Adams (2012).
Congestion Control of Multipath Parallel Transmission of Data for Blockchain Applications
Published in IETE Journal of Research, 2022
The working process of the TCP congestion control mechanism and queue management algorithm has been analyzed, and then the essential relationship of parameters in the system model under a stable state has been studied. Packet dropping is detected by three repeated ACKs. TCP ignores the timeout and slow start and only considers the avoidance of congestion.To ensure the simplicity of the model, the halved time interval between the packet dropping and the detection window of the source host is ignored.The bandwidth of the bottleneck link is fully utilized. All source hosts have enough packets to send so that the bottleneck link is always in the state of full load.A packet is discarded by the probability of the RED algorithm, but the queue overflow is not considered.
A review on big data real-time stream processing and its scheduling techniques
Published in International Journal of Parallel, Emergent and Distributed Systems, 2020
Nicoleta Tantalaki, Stavros Souravlas, Manos Roumeliotis
Drizzle [70] is also a topology-aware strategy that uses the communication structure of a DAG. The main goal of this strategy is to decouple processing intervals from coordination intervals used for fault tolerance. To achieve this, the strategy uses a central scheduler that implements micro-batch groups scheduling, all groups at once. This avoids central scheduling bottlenecks and data processing is completely decoupled from scheduling decisions. An adaptive group-size tuning algorithm inspired by TCP congestion control is used. During the execution of a group, counters are used to track the amount of time spent in various parts of the system and a policy analogous to AIMD determines the coordination frequency for a job. The tuples are processed within a range of milliseconds while coordinating functions take place within a range of seconds. Add-on strategies are used to improve performance. Specifically, pre-scheduling is used to let the worker machines track the data dependencies and run the task when dependencies are met. Query optimisation techniques are used to achieve better throughput. The strategy was compared to Spark and Flink in terms of end-to-end processing delay and failure recovery time and it was reported to have 3–4 times lower latency than Spark. Another comparison metric was failure recovery, which is performed 4 times faster compared to Flink.