Research on multicasting in single-hop WDM networks has so far focused on networks based on the Passive Star Coupler (PSC), a broadcast device. It has been shown that the multicasting performance is improved by partitioning multicast transmissions into multiple multicast copies. However, the channel bottleneck of the PSC, which does not allow for spatial wavelength reuse, restricts the multicast performance. In this paper we investigate multicasting in a single-hop WDM network that is based on

more »

... n Arrayed-Waveguide Grating (AWG), a wavelength routing device that allows for spatial wavelength reuse. In our network optical multicasting is enabled by wavelength-insensitive splitters that are attached to the AWG output ports. Multicasts are partitioned among the splitters and each multicast copy is routed to a different splitter by sending it on a different wavelength. We demonstrate that the spatial wavelength resuse in our network significantly improves the throughput-delay performance for multicast traffic. By means of analysis and simulations we also demonstrate that for a typical mix of unicast and multicast traffic the throughput-delay performance is dramatically increased by transmitting multicast packets concurrently with control information in the reservation MAC protocol of our AWG based network. multi-destination traffic load is due to emerging applications, such as teleconferences, multimedia stream distribution, tele-medicine, and distributed games, and is further increased by the placement of content distribution proxies in metro networks. With multicasting, a source node reaches multiple destinations by sending a single multicast data packet, instead of sending multiple unicast packets. Thus, multicasting can significantly increase the efficient resource (transmitter, channel) utilization for multi-destination traffic and can improve the cost effectiveness, which is critical for metro networks. In this paper we focus on single-hop WDM networks, where source and destination communicate directly with each other, without any traffic forwarding by intermediate nodes. Compared to multi-hop networks, single-hop networks have the minimum mean hop distance (unity) and do not waste any bandwidth for data forwarding. Thus, single-hop networks have the potential to provide a higher channel utilization and an improved throughput-delay performance compared to their multi-hop counterparts [8] . Since the mid-1990s multicasting over single-hop WDM networks based on the passive star coupler (PSC) has received considerable interest [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23]. With the emergence of the metro-gap, multicasting over PSC based single-hop WDM networks has received re-newed interest [24], [25], [26], [27], [28], [29], [30]. A key problem with multicasting in PSC based networks is that the larger the multicast size, the more difficult it is to find free receivers at all destination nodes, especially in heavy traffic. As a consequence, multicast transmissions have to be delayed or other transmissions have to be pre-empted, resulting in a decreased throughput-delay performance. To address this problem, the partitioning of a multicast transmission into several subgroups has been proposed [12] . Instead of sending a given multicast packet to all intended receivers at once, an improved throughput-delay performance is achieved by sending multiple copies of the multicast packet; each copy is received by a different destination subgroup. With partitioning each multicast copy requires a smaller number of receivers which are more likely to be free. Also, simultaneously other transmitters can send multicast packets on different wavelengths to other free receivers. Thus, multiple wavelengths are used at any given time, resulting in an increased efficiency. The PSC, however, is a broadcast-andselect device. Thus, each multicast copy is distributed not only to the intended receiver subgroup but to all receivers, which wastes power and bandwidth. Recently, it was shown that partitioning suffers from a channel bottleneck in PSC based single-hop WDM networks [28], [29] . This is due to the fact that partitioning requires more wavelengths. Since the PSC does not allow for spatial wavelength reuse, the number of available wavelength channels is limited and this prevents nodes from taking full advantage of the partitioning. We investigate multicasting in a network that consists of an Arrayed-Waveguide Grating (AWG) and wavelength-insensitive combiners and splitters attached to the AWG input and output ports. As opposed to the PSC, the AWG is a wavelength-routing device. Each multicast copy is routed to a different splitter by sending it on a different wavelength. Thus, a multicast is partitioned among the different splitters, and each multicast copy is received only by nodes which are attached to the respective splitter. The splitters 3 are used to enable optical multicasting and are located at the network periphery. Each multicast packet is not duplicated until it reaches that splitter to which the corresponding receivers are attached. The same wavelength can be spatially reused in order to send other multicast packets to other splitters. Therefore, compared to the PSC, the AWG provides a higher degree of concurrency which in turn improves the throughput-delay performance of the system by means of partitioning and spatial wavelength reuse. To our knowledge this is the first paper to investigate multicasting in a single-hop WDM network that is based on an AWG. This paper is organized as follows. In Section II we discuss the physical properties of the AWG and the basic principles for multicasting over the AWG based network. In Section III we describe the AWG based network architecture. In Section IV we outline the employed Medium Access Control (MAC) protocol. In Section V we study the throughput-delay performance for multicasting in the proposed AWG based network and compare it with the widely studied PSC based networks. This section focuses exclusively on multicast traffic, i.e., all data packets are multicast packets. We re-confirm the benefits of partitioning the multicast transmissions and demonstrate that the AWG based network with its spatial wavelength reuse has the potential to achieve significantly better multicast performance than the PSC based networks. Next, in Section VI we analyze the transmission of a typical mix of unicast and multicast traffic over the AWG based network. For this typical traffic mix scenario we examine the transmission of ¢ ¡ ¤ £ multicast packets concurrently with spreaded control information during the periodic reservation phase of our MAC protocol (thus increasing receiver utilization and multicast throughput), and ¢ ¡ ¥ ¡ ¤ £ unicast packets with spatial wavelength reuse. We summarize our findings in Section VII. II. PRINCIPLES OF MULTICASTING OVER AWG BASED NETWORK In this section we review the key properties of the AWG which enable efficient multicasting in a singlehop WDM network. Without loss of generality we consider an AWG with degree ¦ = 2, i.e., a § © § AWG, see Fig. 1 . In this example, four wavelengths are launched into the upper input port of the AWG. Every second wavelength is routed to the same output port. This period of the wavelength response is called free spectral range (FSR). We use = 2 FSRs of the underlying AWG, each consisting of two wavelengths. A wavelength-insensitive combiner is attached to each AWG input port. Similarly, a wavelength-insensitive splitter is attached to each AWG output port. Each splitter equally distributes all incoming wavelengths to all attached receivers, resulting in splitting loss. Similarly, each combiner suffers from combining loss, as illustrated in Fig. 1 . Several approaches to compensate for these losses and other network feasibility issues are discussed in [31] . In brief, one possible solution is to place erbium-doped fiber amplifiers (EDFAs) between each combiner (splitter) and AWG input (output) port,