4TH INT. TEL. NET. WORKSHOP ON QOS IN MULTISERVICE IP NETWORKS 2008 1 Extension of the Flow-Aware Networking (FAN) architecture to the IP over WDM environment V. L´ opez * , C. C´ ardenas † , J. A. Hern´ andez * , J. Aracil * , M. Gagnaire † * Universidad Aut´ onoma de Madrid, Spain Email:{Victor.Lopez,Jose.Hernandez,Javier.Aracil}@uam.es † ´ Ecole Nationale Sup´ erieure des T´ el´ ecommunications , Paris, France Email:{Cesar.Cardenas,Maurice.Gagnaire}@enst.fr Abstract—Backbone networks are migrating to IP over WDM architectures. In such multi-layer network configurations, it is necessary to combine efficiently the resources of both layers in order to provide enhanced Quality of Service (QoS) to the end- users. In the context of existing IP networks, Flow- Aware Networking (FAN) has been proposed in order to provide QoS guarantees to multiplexed IP flows within an IP router. FAN is based on implicit admis- sion control and per-flow scheduling. In this paper, we propose a new node architecture that extends the FAN concept to IP over WDM overlayed networks in which both optical and electronic resources are available. Three different policies are introduced to decide on which criteria an IP flow arriving at a node must be bifurcated from the standard FAN architecture to be forwarded onto a transparent lightpath up to its destination. The performance of the three proposed policies are discussed in terms of goodput and of queueing delay. Index Terms—Quality of Service; FAN; Multi- Layer FAN; Traffic engineering; IP over WDM. I. I NTRODUCTION The provisioning of Quality of Service (QoS) to applications implemented at the end-nodes is one of the key issues in the engineering of the Next-Generation Internet. Besides network re- source overprovisioning, two main approaches for QoS support in IP networks have been proposed in the literature: IntServ and DiffServ. The former is well-known for its lack of scalability due to the soft state of the virtual circuits on which rely this tech- nique [1]. The latter requires a signaling channel and a control plane based on complex algorithms The authors would like to thank the support from the Eu- ropean Union VI Framework Programme e-Photon/ONe+ Net- work of Excellence (FP6-IST-027497). They are also thankful to Abdesselem Kortebi from France Telecom R&D for his support in the FAN implementation and to Mrs Sara Oueslati from France Telecom R&D for our fruitful discussions. to address packet marking and metering. Such a mechanisms lead to an expensive approach for QoS provisioning. In this context, a new approach, called Flow-Aware Networking (FAN [2], [3], [4], [5]) has been proposed as a promising technology to manage congestion control in IP networks and to provide QoS to applications. Essentially, FAN operates at packet level and im- plicitly distinguishes between two types of flows: streaming (or priority flows), and elastic (or non- priority flows). Streaming flows typically refer to voice or video applications (UDP), whereas elas- tic flows typically refer to TCP/IP sessions. The FAN architecture is designed in order to meet two main objectives: (1) minimize the queueing delay suffered by streaming flows in the routers; and (2) intend to provide a minimum fair rate to the elastic flows. When FAN cannot satisfy these minimum requirements, it rejects incoming flows. By using such an admission control policy, FAN keeps into service the already admitted flows, thus assuring a minimal QoS under overloaded conditions. In a first approach, FAN was conceived to oper- ate at the IP level without any information about the underlying layers. With the current technology trends, network operators are gradually migrating to an IP over WDM paradigm, mainly to benefit of a larger transmission capacity on each opti- cal fiber. Unlike current protocol stacks such as IP/ATM/SDH, the IP and WDM layers remain relatively independent. This is the reason why we propose an adaptation of the FAN concept to IP/WDM architectures, and by extension to multi-layer capable routers including optical and electronic switching [6]. We define the concept of Multi-layer FAN (MFAN) as a router architecture enabling to handle optical resources provided by WDM technology