Telia are currently in the process of deploying its broadband IP-access. This access designed according to many requirements such as those on bandwidth, cost, quality and security. Also, it is required that the customer is able to choose a telecom operator upon establishing a connection.

For Telia, broadband accesses will mainly be based on ADSL and Ethernet technologies, both which are cost effective in many situations and have very high bandwidths. Quality of Service will be based on priorities, that is the IP DiffServ model, traffic policing and traffic shaping. Security will be accomplished with Virtual LAN and login. These are all traditional and well-known technologies. The ability to choose a network operator can be a harder problem.

The Point-to-Point Protocol, PPP, has been used for many years in network accesses and facilitates Authentication, Authorization and Accounting, AAA, e.g. with widespread Remote Dial-In User Service, RADIUS, efficient IP address assignment and a possibility to choose telecom operator upon connecting. However, PPP requires a one-to-one relationship between communicating hosts while Ethernet is not a one-to-one serial link. PPP over Ethernet (PPPoE) is defined in [1] and provides a way of running PPP on Ethernet. There is a way of running PPP over ADSL. Hence Telia is investigating the possibility to use PPP in its broadband accesses to facilitate the choice of operator.

When using PPP, Quality of Service can be provided through ATM Permanent Virtual Circuit (PVC) services when using ADSL, but for Ethernet with PPPoE the QoS issue remains open. PPPoE has not yet addressed QoS. The IP Type of Service field or Differentiated Services model offers QoS in the IP network as a whole. PPP has built in priorities, as has Ethernet with the 802.1p standard.

In order to investigate the possibilities of introducing QoS support in PPPoE this master project was started [2]. Its preliminary title was "Quality of Service in Point-to-Point Protocol over Ethernet (PPPoE)".

The aim is to investigate PPPoE from a QoS perspective. To clarify PPPoE’s QoS properties. And to answer questions about its behavior and performance under high traffic loads.

This section defines the model for the project and the tasks to be completed in the project.

Model

This section defines a general model of the problem. Telia’s broadband IP-access is, as stated above, based on ADSL and Ethernet technologies. In both cases the customer’s equipment are connected with a switched Ethernet LAN to/in the Customer Premises Equipment, CPE, and from there carried via an access network to an IP-access node. The IP-access node is basically a router where traffic shaping and AAA is performed. The customer could connect for example Set Top Boxes, STBs, IP Telephones, IPT, and ordinary Personal Computers, PCs, to his/her CPE. The above and the following is based on Figure 1, describing the broadband IP-access.

The access network between CPEs and the IP-access node is tree structured. Either Ethernet or ADSL can be used and then the CPE will be an Ethernet switch or a PC, or an ADSL modem respectively. In the ADSL case one or more DSLAMs connect to the access node. In the Ethernet case, a switched Ethernet with 802.1Q Virtual LAN (VLAN) and, typically, two Ethernet switches are normally used on any connection between a CPE and the access node.

Capacities in the access tree are carefully planned. Every VLAN consist of the customers equipment, their corresponding CPE and the IP-access node to ensure that customers cannot intervene other customers traffic unless passing through the policing in the access node. The same VLAN functionality is easily achieved in the ADSL case as each ADSL subscriber is separated with ATM PVCs to the IP-access node.

 

 


Figure 1. Broadband IP-access with Ethernet and ADSL. Ethernet bandwidths are just examples.

When PPP is used, each customer equipment uses a protocol stack with IP on top of PPP on top of PPPoE on top of Ethernet, as illustrated in Figure 2. In order to pass QoS information, i.e. DiffServ/TOS information, from the IP-layer to Ethernet’s "p-bits", this information will have to pass through, and hence be supported by, PPPoE. Ethernet frames can then receive their fair share of QoS when transported to the IP-access node.


Figure 2. Protocol stacks.

The IP-access node can with the help of RADIUS determine how to handle logins and traffic from different users.

Tasks

Assumptions

This project concerns only the broadband IP-access based solely on Ethernet.

This project will not be concerned with the effects that different design approaches, in designing the capacity of Ethernet accesses, has on QoS.

The scope may be limited further once the understanding of tasks involved is deeper.

The methods used in this project are both practical and theoretical. The theoretical parts will be investigated by learning more about the architecture, involved protocols and the relations between them in this specific case. Hence, building up an understanding of the system as whole. This is achieved with the Literature Study.

The more practical parts of the project, id est when studying PPPoEs behavior under high traffic loads, will be carried out with lab measurements. A computer lab with an experimental setup of the IP-access will be at the disposal of the project. With the help of traffic generators, traffic analysers and a packet delayer (possibly modified for this purpose) the effects of concern here can be studied.

1. Introduction
   About broadband accesses, PPP, Telia IP-network and broadband IP-access, background, motivation, motivation of PPP usage. A little about what has been done in the area.
   Report outline…
2. An overview of PPP and PPPoE
   About PPP and PPPoE in general
3. QoS in IP, PPP, PPPoE and Ethernet
   About priority/DiffServ schemes in IP, PPP, PPPoE (earlier proposed methods for QoS and point out lack of…), and Ethernet.
4. Problem Definition and Model
   Definition of problem, description of model.

 

5. Alternatives for QoS in PPPoE
   Theoretical discussion on how to provide QoS in PPP/PPPoE.
6. Multiplexing of different QoS-classes (onto one PPPoE session)
   What to do about this. Or how to avoid it altogether.

 

7. A proposal for provisioning of QoS in PPPoE
   How to provide QoS in this stack. What advantages this proposal has, disadvantages etc implications on…

 

8. Lab measurements
   Aim of tests. PPPoE’s reaction to loss, delay, and reordering and performance consequences Test Description, how tests were carried out
9. Results
   The results are presented along with implications

 

10. Implementation of QoS support in PPPoE
    Description of implementation on the Linux open source platform. Code possibly in an appendix.
11. Test
    Basic tests of implementation

 

12. Conclusions
    Sum up (and wrap up) conclusions
13. Future work
    Unfinished work, if any. Recommendations for further studies.

 

14. References
15. Appendices
    For sections necessary for the report to be comprehensive that, if not placed here, would break its account apart.

1. Mamakos, L. Et al. 1999. A Method for Transmitting PPP Over Ethernet (PPPoE). Informational RFC2516, The Internet Engineering Task Force.
2. Johansson, Fredrik. 2000. Beskrivning av examensarbete - QoS i PPPoE. Telia Research AB, Stockholm.