Preemption Window mechanism for efficient QoS support in E-OBS network architecture Davide Careglio * , Miroslaw Klinkowski * , † , Josep Sol´ e-Pareta * * Advanced Broadband Communication Center Universitat Polit` ecnica de Catalunya, Barcelona, Catalunya, 08034 Spain Email: {careglio, mklinkow, pareta}@ac.upc.edu † National Institute of Telecommunications Warsaw, 04-894 Poland Abstract—This paper focuses on the problem of quality of service (QoS) provisioning in optical burst switching (OBS) networks. OBS is a promising photonic network technology aiming at efficient transport of IP traffic by means of statistical multiplexing. The lack of optical memories, however, makes this operation quite complicated. Problems such as unfairness in access to the shared transmission resources, facility in adopting alternative and backup routing, scheduling complexity and so on arise in the conventional OBS architecture. In [1] we proposed the offset-time emulated OBS (E-OBS) architecture, which overcomes all these drawbacks by means of distributed provisioning of the offset time in core nodes. Nonetheless it is still difficult to guarantee a certain level of service quality. Burst preemption mechanism, which, alongside with offset-time differentiation, was proven to be the most effective technique for QoS provisioning in OBS networks. The general drawback of any burst preemption- based mechanism is that, in case of successful preemption, either the resources reserved for the preempted bursts on outgoing path are wasted or an additional signaling procedure should be carried out in order to release them. In order to avoid wasted resources reservation, in [2] we proposed the Preemption Window (PW) mechanism which enhances the E-OBS for efficient QoS support. In this paper we evaluate exhaustively the performance of the resulting architecture showing all its advantageous with respect to other solutions. I. I NTRODUCTION Optical burst switching (OBS) is a promising solution for reducing the gap between switching and transmission speeds in future networks [3]. Packets coming from client networks are aggregated and assembled into optical data bursts in the edge nodes of an OBS network. A burst control packet (BCP) is transmitted through a dedicated control channel and delivered prior to the data burst (the so called offset-time). In this way the electronic controller of an intermediate (core) node has enough time both to reserve a wavelength on its output link, usually for the duration time of the incoming burst, and to reconfigure dynamically the switching matrix. The output wavelength is released for other connections when the burst transmission is finished in the node. Such a temporary utilization of wavelengths allows for higher resource utiliza- tion as well as for better adaptation to highly variable input traffic in comparison to optical circuit-switching networks. Moreover the aggregation of data packets helps to overcome the fast processing and switching requirements of optical packet switching (OPS) technology. In fact, OBS allows using state-of-the-art switching elements [4]. There are two distinct signalling architectures considered for OBS networks. The first one is based on a connection-oriented signalling protocol which performs end-to-end resources reser- vation with acknowledgment in so called two-way reservation mode. The other exploits a connection-less signalling protocol which allocates the resources on-the-fly, a while before the burst arrival, in a one-way reservation mode1. Since the prob- lem of the two-way reservation signalling concerns the latency due to the connection establishment process such architectures are less interesting for long-haul network applications due to the large latency and are not addressed in this paper. The one-way reservation signalling that can operate effec- tively in large distance OBS networks performs according to a statistical multiplexing paradigm; hence it encounters the problem of burst contention inside the network. Indeed, when a burst control packet enters a node in order to perform the wavelength reservation for its data burst, it may happen that the requested resources are not available at the output link and the burst has to be dropped. The lack of optical random access memories complicates the resolution of burst contention in optical networks. To alleviate this problem several mechanisms based on wavelength conversion, deflection routing and fibre delay line (FDL) buffering together with dedicated burst scheduling algorithms have been proposed. From the very beginning, there were two distinct control architectures considered for OBS networks [3]. The difference between them comes from different management of offset times. A conventional OBS (C-OBS) introduces the offset time in soft-way by delaying the transmission of burst with respect to the BCP in the edge node. At each core node, the offset time decreases by the time the BCP spends in the switch controller. Another idea for an OBS operation comes from OPS world and it intends to emulate the offset time by means of an additional fiber delay unit (FDU) introduced in the data path at the input port of the core node in the so called offset time emulated OBS (E-OBS) architecture. FDU delays the arrival of the burst with respect to the arrival of its BCP and in such hard-way it introduces the offset-time. Although C-OBS has attracted lots of attention we highlighted in [1] that problems such as unfairness in access to the shared transmission resources,