Congestion Control in computing networks: problems and trends

Source: Internet
Author: User

Abstract:

First, describe cheaper memory. High-speed links and higher-speed processors will solve computer network congestion. Then, we recommend a simple congestion definition and various congestion modes for supply and demand. It makes congestion control difficult to discuss and present architecture decisions that affect the design of the congestion mode. There are different methods to address long-term, medium-term, and short-term congestion issues.

 

Introduction

Congestion Control is about how to allocate resources to obtain an acceptable performance level when the demand exceeds or is close to the network resource capacity. Network Resources include linked loans, buffer space (content), and intermediate node processor capabilities. Although resource allocation is important even when the load is low, as the load increases, the fairness and lowOverheadMore importantly, if there is no proper congestion control mechanism, the network throughput will be greatly reduced under heavy loads.

 

Myth About congestion control

Congestion occurs when the demand is greater than the available resources. Therefore, if the resources are no longer expensive, the congestion problem will be solved. This is the myth below:

1.Congestion is caused by a lack of space and can be solved with low memory, because there is unlimited memory.

2.Congestion is caused by low-speed links and is solved with the popularization of high-speed links.

3.Congestion is caused by a slow processor and is solved as the processor speed increases.

In fact, even if the above statements do not have proper protocol control, the above solution will not solve the congestion, but lead to more congestion, thus reducing performance. The following is a detailed explanation.

1.Large buffer space cannot solve the congestion problem. Some studies have found that vswitches with infinite buffer sizes are congested the same as those with limited buffer sizes. A vswitch with a limited buffer zone may lose packets, but a vswitch with an infinite buffer zone will retransmit data because the data buffer is too long and the application layer considers the data to be invalid. In short, the infinite buffer is actually harmful, mainly because it may lead to invalid use of network resources.

2.The use of high-speed links does not solve the congestion problem, but increases the need for congestion control. Because all high-speed networks are connected to low-speed networks, and as the number of high-speed networks increases, the low-speed networks will not automatically disappear. This requires the application layer design protocol to ensure that performance will not be reduced.

3.The introduction of high-speed processors cannot solve the congestion problem, because it is the same as the link rate, which will also lead to the problem of mismatch of the congestion rate.

4.This problem cannot be solved even if all network resources are at the same speed. This is because the network cannot be properly configured anywhere and there is no mismatch.

Therefore, congestion is a dynamic problem and cannot be solved through static means. This requires protocol design to save the network under congestion. The expansion of high-speed networks leads to more unbalanced networks due to congestion, and the packet loss caused by the lack of buffer zone is a symptom but not a cause of congestion.

 

Definition and solution of congestion problems

In a computer network, if the total number of resources required for a period of time is smaller than the amount of available resources, it will lead to simple congestion. There are many resources in the computer network, such as buffers, link bandwidth, processors, and servers. If the buffer space at an interval is insufficient, packet loss occurs. To put it simply, if the total traffic expected to enter the link in a short period of time is greater than the bandwidth, the link is congested.

There are two types of congestion modes:

1.Dynamically add available resources

2.Dynamic Demand Reduction

Resource Creation Mode

This mode is determined by dynamic resource reconfiguration. For example:

1.Add dial-up connections at high usage

2.Increase the power of satellite links to increase bandwidth

3.Consider using a low-load non-optimized network for extra Data Transmission

In all the above modes, resource users do not need to be notified, they do not even know the network congestion, and the network itself is responsible for the congestion problem.

Demand Reduction Model

This usually requires the resource user to know the load status to adjust the traffic. Three basic modes:

1.Service Denial mode: New sessions cannot be established during congestion. Connection-oriented computer networks use similar models to establish sessions on intermediate nodes.

2.Service downgrade mode: This mode reduces the load for all users (existing or newly added. Dynamic window is the embodiment of this method.

3.Plan mode: This mode allows users to plan resources and ensure that the resources are smaller than the capacity. It must be noted that this mode is only a special case of the downgrade mode.

In a non-connection network, the establishment of a session does not require the consent of the intermediate node. Therefore, the denial of service mode is generally invalid, but the service downgrade mode and plan mode are used.

 

Feedback and Control

All of the above models usually measure the network load first and then take remedial measures. The first part is feedback, followed by control. Feedback usually occurs from congested resources to one or more control points. In Demand Reduction Mode, control points are generally data sources, while in creation mode, control points may be intermediate nodes.

 

Feedback mode

1.Feed back the message. A congested resource point generates a separate message and sends it to the control point. If the source receives the message, it reduces the load or increases the load. The cost is extra load.

2.Feedback in routing messages:Each intermediate node sends load information to a neighbor.

3.Inject more traffic: Back Pressure

4.Test package: Send a test package to obtain the load status.

5.The feedback field in the subfolder: the feedback package is included in the reverse data instead of generating the feedback package separately.

 

Control Point Location

1.Transport Layer:Communication traffic is generated by the terminal system, so the best control point is at the transmission layer. For example, dynamic window mechanism.

2.Network access layer: it is like a traffic light indicating the entrance of a highway. Access Control is added at the network layer to restrict traffic from entering the network when congestion occurs.

3.Network Layer: adds congestion control measures on routers and netswitches.

4.Data link layer: The traffic control mechanism is used on the data link layer of each hop.

 

Why is the problem difficult?

Although there are many proposals on congestion control, at least this research has been conducted20For many years, there are two main reasons. First, it is difficult to meet the needs of the congestion control mode to achieve a satisfactory solution. Second, there are several network policies that affect the design of the congestion mode. Therefore, a mode works well in one structure, but not necessarily in other structures. The following are some questions about the second point.

1.The mode must be low-Overhead: This requires no additional overhead for control, which is why it is unreasonable to report messages separately. Some people suggest sending feedback only when the load is low, and the lack of feedback means high load by default.

2.The mode must be fair: fairness is very important when the load is high. The important question is how to rationally allocate resources.

3.The mode must be responsive: the available network resources change. As nodes and links increase or decrease, the available capacity increases or decreases. When users start or stop, the demand increases or decreases. The congestion control mode should be able to dynamically match the demand and available capacity. Therefore, when the available capacity increases, the user needs to be notified to increase the demand, and vice versa. The demand curve should be closely related to the capacity.

4.The mode must work in a poor network environment:

5.

Contact Us

The content source of this page is from Internet, which doesn't represent Alibaba Cloud's opinion; products and services mentioned on that page don't have any relationship with Alibaba Cloud. If the content of the page makes you feel confusing, please write us an email, we will handle the problem within 5 days after receiving your email.

If you find any instances of plagiarism from the community, please send an email to: info-contact@alibabacloud.com and provide relevant evidence. A staff member will contact you within 5 working days.

A Free Trial That Lets You Build Big!

Start building with 50+ products and up to 12 months usage for Elastic Compute Service

  • Sales Support

    1 on 1 presale consultation

  • After-Sales Support

    24/7 Technical Support 6 Free Tickets per Quarter Faster Response

  • Alibaba Cloud offers highly flexible support services tailored to meet your exact needs.