There is a desire among telecom operators to provide communication services in IP networks that meet strong quality-of-service requirements. Consequently, an operator needs to develop efficient tools for monitoring and control of relevant performance parameters. Besides to have a good knowledge about the behaviour of the network for operational purposes, it has also become increasingly important to verify whether service-level agreements are fulfilled or not.
This paper is focused on monitoring of QoS parameters in IP based networks, especially Virtual Private Networks (VPNs). The technique to provide logically private domains within the traditional telephony network has been used for some time now. Since the Internet Protocol today is ubiquitous also in public telecommunications, IP-based VPNs are considered to become an important method for providing secure and reliable telecom services.
The purpose is to present an architecture for performance monitoring in IP networks, with a special focus on VPNs. The proposed framework is based on embedded monitoring, where packets are inserted between blocks of user traffic, which is different from the prevailing methods used in most of today's monitoring tools and systems.
Section 2 gives a background and related works and the general architecture is outlined in section 3. A prototype test implementation in linux-based routers is described in section 4 followed by an evaluation of monitoring packets based on measurement data in section 5. Finally, ways to apply the methods to other cases than VPNs, further development of the method are discussed. The paper is a result of a joint project between the Department of Teleinformatics at KTH and Telia ProSoft.
… ..
In this section we propose an architecture for embedded performance monitoring in IP networks, especially applied to virtual private networks. The method is based on dedicated monitoring packets, which experience the same conditions as the user traffic, and monitoring functions that are an integrated part of the network elements.
The basic idea is to develop an appropriate infrastructure for monitoring of network performance parameters in IP networks. We believe that measurements and monitoring functions have to be determined by the operator's policy and objective for performance management and adjusted to the type of services that are offered. There is no meaning to carry out extensive performance monitoring for its own sake.
We have chosen the emerging IP-based virtual private networks as the target for this study since it represents a case where more elaborated performance monitoring is motivated. As seen in fig. 1 we assume a topology with a core network surrounded by provider edge nodes and customer edge nodes. Besides the obvious need for an operator to be well informed of the behaviour of its network as a whole, monitoring of service-level agreements and quality-of-service has become on important part of an operator's responsibility, possibly supported by customer-based network management systems. Furthermore, a powerful real-time monitoring system, capable of reflecting the actual performance of the network, could also provide support for dynamic capacity allocation functions.
Virtual private networks can be implemented in different ways. In router-based networks so called tunnels are created by means of overlay point-to-point connections, using for example generic route encapsulation or IPSec. Multi Protocol Label Switching (MPLS) promises to provide a more flexible and scalable framework for VPNs based upon ATM switches or a mixed environment with routers and switches.
(Fig. 1: Virtual private networks that are defined between provider edge nodes (PE) or between customer edge nodes (CE) use a common core network.)
… .
The purpose of embedded monitoring is to measure relevant network performance parameters based on the actual user traffic. These dedicated monitoring packets, OAM packets, are inserted between blocks of ordinary data packets as shown in fig. 2. The sending node generates monitoring packets that convey OAM information between every N user packets on average. The receiving node detects the monitoring packets through a unique protocol number, adds information and returns them to the originating node. Processing, storage and analysis may be carried out by dedicated servers for the entire network.
(Fig 2: Two OAM packets enclose an OAM block that consists of N user packets on the average.)
(A format of the OAM packet is proposed)
Using the proposed method and format it is possible to obtain: