Datagram Congestion Control Protocol Y. Dong, Ed.
Internet-Draft C. Liu, Ed.
Intended status: Informational Nanjing Univ. of Posts and Telecom.
Expires: August 19, 2016 February 16, 2016

A Dynamic Service Class Mapping Scheme for Different QoS Domains Using Flow Aggregation


This document addresses the issue of provisioning end-to-end Quality of Service (QoS) for multimedia services over heterogeneous networks and introduces a parametric model by using network calculus theory for QoS class mapping between different QoS domains. Then a QoS Mapping Scheme based on Flow Aggregation (QMS-FAG) is proposed in this document to mitigate the information loss problem due to mapping between QoS domains with different granularity of QoS class and to provide efficient network resources utilization by considering user's Quality of Experience (QoE). In QMS-FAG, the QoS requirements of service flows are indicated by a unique FAG identifier which is described in a service flow map of QoS parameters. With FAG identifier and mapping executors sitting at the border of different QoS domains, QMS-FAG allows smooth QoS class mapping between networks with different granularity of QoS class. Both numerical analysis and simulation studies are given to demonstrate the efficiency of the proposed method.

Status of This Memo

This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.

Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at

Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."

This Internet-Draft will expire on August 19, 2016.

Copyright Notice

Copyright (c) 2016 IETF Trust and the persons identified as the document authors. All rights reserved.

This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents ( in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.

Table of Contents

1. Introduction

This document proposes a unified QoS Mapping Scheme based on Flow AGgregation (QMS-FAG) to provide better end-to-end QoS over heterogeneous networks. Different from previous efforts, the aim of the proposed method is to provide better flow services over heterogeneous networks. We aim to contribute to the ongoing research by proposing a QoS mapping scheme, based on network QoS requirements and users' QoE. The proposed method has several advantages: (1) it considers the asymmetrical problem between fine and coarse grained QoS domains (Normally the fine grained QoS domain has more/finer QoS classes than the coarse grained QoS domain); (2) it considers QoE and can improve users' experience by maximizing the utilization of network resources with flexible QoS class mapping; (3) it does not need a mapping table.

1.1. Requirements Language

The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in RFC 2119 [RFC2119].

2. QMS-FAG Development

Previous studies on QoS class mapping between different networking technologies can be roughly classified into two categories: the function based methods [2][3] and mapping-table based ones [4][5]. The first category that translates between the QoS parameters of heterogeneous networks is complex, in which the effective design of functions can highly affect the end-to-end QoS. The second category that established mapping tables consisting of many QoS class pairs can cause the i nformation loss due to mapping between QoS domains with different granularity of QoS class. One shortcoming of current approaches is incapable of utilizing network resources efficiently because of not considering users' QoE in the QoS mapping process.

For ease of analysis, let us define the following variables:

2.1. QoS and QoS Class Models

Based on [6] and our finding, each of QoS parameters can be parameterized by a real number (Please see Appendix A for details). Assuming that the QoS value is ranked in order of importance in this paper, the most important one has the minimal value and the least important has the maximal value. Each of QoS requirements can then be represented by a real number and the overall QoS requirements can be represented by a vector consisting of corresponding QoS parameters. Formally, we specify the overall N QoS requirements of i-th service by vector Pvi, as follows:

where i is an integer that represents index of service,pvi is in RN represents the value of n-th QoS requirement located in RN space for i-th service. RN denotes an N-dimensional real number Euclidian space which is consisted of QoS parameters.

Because each QoS class has a sub-space in N-dimensional space, we use a pair value (PWhu,PWhl) specification in our paper, which will allow us to define range representation with acceptable QoS regions (PWhl≤Pvi≤PWhu) and unacceptable QoS regions (Pvi<PWhl) of QoS class h in network W with proper normalization of QoS parameters (see Appendix A for details).Pvi<PWhl indicates the level of QoS is below the acceptable lower boundary, with which the traffic should be arranged for the lower class level or refused to transmit. For the case of Pvi>PWhu, it indicates that the traffic with Pvi should be arranged for a higher class level.PWhl and PWhu are the lower and upper boundaries of QoS class sub-space in N-dimensional space, respectively, whose definitions are similar to Pvi, where h=1,...,HW;HW is the number of QoS classes in network W.

QoS influences user's QoE, which is vital for the success of multimedia services. Furthermore, QoE is also influenced by the human factors that often are independent of the service type [7]. As a result, different users of the multimedia service have different tolerance for adjusting QoS level. For some users, when enjoying a live TV program via web (such as a football match), they probably prefer to degrade their QoS level rather than to be denied access directly. Therefore, users' QoE should be considered in QoS class mapping schemes to increase the number of satisfied users in heavy traffic load.

Depending on the individual human perception, it is somewhat difficult to give a precise objective metric and objective estimation method for QoE [8]. This paper will not concentrate on how to estimate QoE or map between QoE and QoS, which has been a hot research topic of many other works [8][9].

We use the QoE model proposed in [9] to obtain a mean opinion score (MOS) to rate QoE level and modify the model by substituting sender bitrate (SBR) with bandwidth and block error rate (BLER) with packet loss rate caused by delay and link errors. In the modified model mean burst length (MBL) and content type (CT) have constant values of 2.5 and 0.1, respectively, which are typical values in [9] (For details see Appendix B). The values of the coefficients of the modified model are the same as the values of the model proposed in [9]. In this paper, by dynamically adjusting the QoS parameter values within threshold, we present an empirical QoS class mapping method with QoE to demonstrate the feasibility of the proposed method.

2.2. Flow Aggregation Concept

A flow aggregation (FAG) is defined in this work as a set of flows with similar QoS requirements represented by a conjunction of a set of F QoS parameters P=[p1,p2,...,pF], each associated with a QoS constraint, that can be specified by a range representation with acceptable and unacceptable QoS regions. We assume that the QoS requirements of a service flow can be expressed by a vector in a multi-dimensional space of relevant QoS parameters, and then define this multidimensional space as a service flow map. Each FAG has a unique identifier that can be described by the QoS information on a service flow map.

The FAG is different from QoS class defined by global standardization organizations in the following aspects: 1) its granularity can be established on the fly according to QoS requirements of services and reflects natural muster in QoS characteristic space, and is not connected with any of the predefined QoS classes; 2) it provides a bridge with a flexible granularity for consistent mapping between fine and coarse grained QoS classes in order to mitigate the information loss problem, whose efficacy will be demonstrated by numerical analysis in Section V.

2.3. An Overlay Network Paradigm

In this section, we describe an overlay network paradigm based on the scenario illustrated in Fig. 1. From the viewpoint of providing end-to-end QoS guarantees, the process of QoS mapping can be imagined as a virtual plane of QoS mapping above the traditional layers. This plane of QoS mapping is a collection of virtual nodes connected together by a set of virtual links to form a large virtual domain, which is essentially a subset of the underlying network topology. Each virtual node is a logical abstraction of a particular physical node that processes QoS mapping. A virtual link spans over a path in the physical network and includes a portion of the networking resources. By allowing multiple networks to have different QoS domains to map QoS in the plane of QoS mapping, users in two ends construct a virtual end-to-end path and are provided end-to-end QoS guarantees across different QoS domains, as illustrated in Fig. 1.

In Fig. 1, the proposed Mapping Evaluator (ME) entity sits on a gateway/router at the edge of two different QoS domains, aiming to classify each service according to QoS requirements. Whenever ME receives a service, it generates a corresponding FAG with P according to QoS requirements of the service by a clustering algorithm, such as evolutionary algorithm, and labels the FAG with a unique FAG identifier. Then ME puts the FAG into the corresponding queue with the same priority value. According to available network resources, ME determines appropriate QoS class mapping between current and new networks for the FAG by the proposed QMS-FAG.

2.4. A Typical Scenario of QoS Class Mapping over Heterogeneous Networks

In this section, we depict a typical scenario of QoS class mapping over heterogeneous networks.

A typical scenario of QoS class mapping over heterogeneous networks is shown below.

          Nwk A -----R1----- Nwk B -----R2----- Nwk C

Figure 1

As illustrated in figure above, we consider a scenario of three interconnected networks (Nwk A, Nwk B and Nwk C) connected by two gateways/routers (R1 and R2). Assume Nwk A and Nwk C are 3G UMTS networks and Nwk B is a wireline IP-based Diffserv network. Since audio conferencing is a typical multimedia service requiring strict QoS requirements to set priorities at flow\packet level, here we assume that the audio conferencing service is implemented between user X and user Y. In source network (Nwk A), an appropriate QoS class queue is assigned to audio according to QoS requirements. For traditional QoS mapping, the QoS class mapping table is preset in the gateway/router that sits at the boundary of two different QoS domains and the audio conferencing service belongs to a certain QoS class of current QoS domain. Whenever the gateway/router receives an audio conferencing service, it determines an appropriate QoS class mapping between current and new network according to the mapping table for this audio conferencing service.

2.5. MOS Value for Video Service

The MOS value for video service is computed as follows [10]:

where, SBR is sender bitrate, BLER is block error rate in 3G/UMTS networks, MBL is mean burst length, CT is content type of the video service.

Coefficients in (2) are:

a1 a2 a3 a4 a5 a6 CT MBL
3.9560 0.0919 -5.8497 0.9844 0.1028 -0.236 0.1 0.25

3. QMS-FAG Description

3.1. Parametric Model

The proposed scheme can automatically map the FAG to the appropriate class that has QoS resource by adjusting QoS requirements. An attractive feature of the dynamic QoS class mapping is that the method considers the QoE of end users by which the ME adjusts the QoS requirements of FAG under the condition of available QoS resources.

Consider a network session being set up over the heterogeneous networks consists of MEs S1,S2,...,Sk, the set of FAGs that will be transmitted into the next network can be described as

  • X = {x1,...,xm,...,xM}, m=1,2,...,M (3)

where xm represents the m-th FAGs, M is the number of FAG in an ME.

By a similar description to QoS class, xm can be described as

  • FWm = [Pm1,...PmN], m=1,2,...,M (4)

where FWm denotes QoS description of xm in network W, Pmn(n=1,2,...,N) represents the n-th QoS requirement of xm.

In mapping process, the ME will map xm to class y described as below by the function Phi(for RN->F, then x->Phi(x)). This function can be derived as [2]

  • ch=W*(||FWm-PWh||), for all h=1,2,...,HW (5)

where ch is the order of QoS class mapped according to PWm, W=[w1,w2,...,wHW] is a weighting array which is used to describe the characteristics of multimedia service, satisfied with w1+w2+...+wHW=1 and often gained based on the experience. But the computation is different in the two following cases: lower and higher traffic load cases.

3.2. Procedure of QMS-FAG

Computation is different in the two following cases: lower and higher traffic load cases.At lower traffic load, QoS class y can be derived as

  • y = {k|ck=minh{ch}}, for all h=1,2,...,HW (6)

where k is the order of QoS class and ck is the minimum value among all QoS classes ch, y is the order of QoS class adjusted according to available network resources. Here, if one QoS class has a smaller order value, the class's FAG has a better chance to transmit earlier.

At higher traffic load, the process is as follows:

The QMS-FAG scheme at higher traffic load is described in Algorithm 1.

/*Algorithm 1: The QMS-FAG scheme */
1. The QoS class level is decreased by one
2. y is recomputed according to equations (6) based on the QoS requirements adjusted   
3. If network resources for the decreased QoS class are still not enough to transmit 
	this FAG, go back to step 1)
4. If the MOS value is still OK (above a preset threshold MOSth) for end users, 
	based on Equ. (2), then, this FAG is transmitted; otherwise, 
	the FAG is rejected.  
6. The process is stopped.

On the whole, the algorithmic steps of QoS class mapping are as follows:

1)If the network resource allows, ck is mapped to QoS class y based on equations (2)

2)If the network resource is not allowed, with the process in the case of higher traffic load, the ME gradually reduces the order of QoS class for xm until xm is transmitted with a lower order of QoS class, or is rejected if no appropriate mapping is available (assuming the lower the order of QoS class, the lower its priority).

4. Acknowledgements

The authors would like to acknowledge feedback and discussions on service class mapping scheme for QoS with a wide range of people, including members of the Wireless Communication Research Group and the End-to-End Research Group. Thanks are given to the National Natural Science Foundation of China (No.61271233, No.60972038), the Ministry of Education (China) Ph.D. Programs Foundation (No.20103223110001), the Research Culture Funds of Anhui Normal University (No.2013xmpy10) and Jiangsu Province Postgraduate Innovative Research Plan (No.CXZZ11_0396) for their financial support.

5. IANA Considerations

There are no IANA actions required for this document.

6. Security Considerations

All drafts are required to have a security considerations section.

7. References

7.1. Normative References

[1] Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, DOI 10.17487/RFC2119, March 1997.

7.2. Informative References

[2] Misun Ryu, Youngmin Kim, Hongshik Park, , "Systematic QoS Class Mapping Framework over Multiple Heterogeneous Networks", September 2008.
[3] Wang Zhenhua, Sun Qiong, Huang Xiaohong, Ma Yan, , "IPv6 end-to-end QoS provision for heterogeneous networks using flow label", 2010.
[4] Lin Fu, Fei Pei, Zhang Dengyi, Li Wenhai, , "Quality of service support for event detection in wireless sensor networks", 2011.
[5] Ben Hamza Nejd, Rekhis Slim, Boudriga Noureddine, , "Cooperative architecture for QoS management in wireless 4G networks", 2011.
[6] Klara Nahrstedt, Jonathan Smith, , "The QoS Broker", 1995.
[7] Wanmin Wu, Md Arefin, Raoul Rivas, Klara Nahrstedt, , "Quality of Experience in Distributed Multimedia Environments: Towards a Theoretical Framework", October 2009.
[8] Kye-Hwan Lee, Son Tran Trong, Bong-Gyun Lee, , "QoS-guaranteed IPTV service provisioning in IEEE 802.11e WLAN-based home network", 2008.
[9] Asiya Khan, Lingfen Sun, Emmanuel Ifeachor, , "QoE Prediction Model and its Application in Video Quality Adaptation Over UMTS Networks", 2012.
[10] Cibin R, Sudheer K P, Chaubey I, , "Sensitivity and identifiability of stream flow generation parameters of the SWAT model", 2010.

Authors' Addresses

Yu-ning Dong (editor) Nanjing Univ. of Posts and Telecom. 66 New Mo-fan-ma-lu Road Nanjing, Gulou 210003 China Phone: +86 15077858011 EMail:
Chun Liu (editor) Nanjing Univ. of Posts and Telecom. 66 New Mo-fan-ma-lu Road Nanjing, Gulou 210003 China Phone: +86 18362930657 EMail: