HSC welcomes all external visitors to this site, especially students and members of the academic community. Please use the comments box at the bottom of each page to record any comments or suggestions for improvement.
Introduction
Given the limiting nature of the communication channel in a communication system, the end user applications (such as browsers, FTP clients, media clients) experience certain effects of the communication channel. These effects, such as delay, throughput, error rates, variation in delay etc. constitute the elements that make up for the user experience or "service".
The teletraffic engineering world was a relatively simpler world in the voice telephony era and very effective techniques of measuring and engineering service quality are in place. In this section we shall consider only the packet data networks and use the term network to mean packet data network.
The goals of QoS Engineering
The key components of engineering "service quality" in any type of communication system are:
1. Service quality
2. Communication network model
3. Source Traffic Model
4. Network Traffic model
The aim of QoS engineering is to define the service quality parameters in a given network, arrive at suitable traffic models that reflect the source traffic as closely as possible and model the communication link behaviour.
ITU-T defines the teletraffic engineering as the iterative process of:
Demand Classification?
Grade of Service definition
Dimensioning and control?
Performance Measurement?
Quality of service engineering encompasses of multiple related fields, each one useful in one of the phases of the teletraffic engineering. An end to end Quality of service in a heterogeneous network is an interesting field of study, albeit, a wide one.
Consitution of a QoS system
A QoS system consists of a collection of entities, each with one or more specialized functions, acting under a common design policy to implement a Quality of Service model within a network. The diagram below shows a sample decomposition of a QoS system. The components within the QoS system exchange information with each other explicitly or implicitly. They also receive and react to external events. Some of them have functions to do with standard network handling which impact QoS and the others implement specific functions to achieve the common QoS policy.
In the diagram above, we have three components, the end-point TCP stack, the network layer concentrator and the link layer frame-builder which, acting together comprise the end-to-end QoS system. Note that the TCP and the link layer frame builder are specialized components with their own specific functions. If the entire network, as a whole, must come together to implement a specific QoS model, then the individual pieces must be controlled so that:
- Individual reaction to external events are complementary.
- Causal reactions to specific events are self-stabilizing and, taken as an aggregate, cause an overall reaction that 'makes sense'.
- The handling of a single service unit is executed in a consistent manner.
One of the major challenges of designing a QoS system is to prove that the above holds. This is especially so when all the applications, including the endpoints are taken into account. One solution is to show that they fall into an existing proven QoS framework, such as Diffserv or Integrated Services. However, this requires that all the critical components in the system are compliant. Consider, for example, the link layer frame-builder, which does not even see full IP packets - clearly the above goal is not practical. To solve this problem, it is possible to ensure that the information available to the link layer frame builder has been processed so that, acting on the basis of this information, this component automatically becomes compliant to the framework. This is part of the challenge of QoS system design.
Subtopics
Fairness?. Discusses definition, technical results, etc. in the theory of fairness
Scheduling?. Article on types of scheduling, i.e. csdp, gps, etc. There is a currently a separate article on gps which can remain.
Traffic modeling? Technical article on modeling internet traffic.
QoS architectures in IP? Discussing QoS architectures and contexts (diffserv, intserv, etc.) in IP
QoS issues in Wireless? Discussing QoS in the special context of wireless networks. Include issues of csdp, mobility, power control, etc.
QoS issues in TCP? Article on what kind of QoS TCP currently supports, integrating QoS schemes with TCP, etc.
QoS for multimedia networks? Article on QoS requirements for multimedia
References
[garg] Rahul Garg, Abhinav Kamra, and Varun Khurana. A game-theoretic approach towards congestion control in communication networks. Computer Communications Review, 32(3):47-61, June 2002.
[Ohsaki] "H. Ohsaki, M. Murata, T. Ushio, and H. Miyahara, A control theoretical analysis of a window-based flow control mechanism in tcp/ip networks, submitted to IEEE Transactions on Control Systems Technology, August 2000."
Categories: QoS
Comments
