This paper describes both the experimental and theoretical evaluation of the cascadability of all optical routers in optical label swapping networks incorporating a multistage wavelength conversion with 2R regeneration. A full description of a novel experimental setup allows the packet-by-packet measurements in systems with up to 16 hops and 10 Gbit/s payload showing 1 dB penalty with 10-12 bit error rate. The conducted simulations on the system test-bed predict router scalability for systems with up to 64 hops.
©2006 Optical Society of America
The currently noted exponential growth of data traffic demands deployment of new techniques for the packet switching based networks. The issue of supporting increasing bandwidth demand and simultaneously reducing the bottleneck caused by processing individual packets at network nodes is still open. In the last few years the paradigm of routing packets directly in the optical layer emerged as a technical approach to supporting the next generation Internet, based on All-Optical Label Swapping techniques (AOLS) [1–4], which combine features of MultiProtocol Label Switching (MPLS), namely traffic engineering with optical packet switching to achieve routing and forwarding of IP packets in the optical domain. In AOLS approach, the routing information (label) and the payload (IP packet) are transmitted at the same wavelength by using complex modulation techniques, such as SubCarrier Multiplexing (SCM) [1,2], bit serial time domain , orthogonal modulations , etc.
The SCM approach exhibits several advantages since payloads and optical labels are conveyed at the same wavelength and thus their subsequent separation is easily achievable by simple optical filtering techniques. The subcarrier optical label can be removed and/or replaced in a more asynchronous manner with respect to the packet payload, unlike in a serial transmission case where such operations are time critical.
Cascadability is a key issue for optical packet routers with label swapping capabilities , since the nodes in a network include both passive and active devices which degrade the signal quality. As a result, they have an impact on the network dimensioning. In this paper we demonstrate and validate the multihop behavior of an all-optical label router which consists of filtering, label processing, wavelength conversion, label rewriting and routing stages.
The organization of the paper is as follows. Section 2 describes the subsystems of the all-optical label swapping router including the multistage wavelength conversion. Section 3 demonstrates the network dimensioning process by simulating the main node subsystems, while section 4 presents experimental results on the performance of the router. Section 5 describes the experiment methodology and the obtained results and finally section 6 summarizes the paper.
2. Description of multistage 2R regeneration all-optical swapping node
The architecture of the all-optical label swapping router is depicted in Fig. 1. The router is composed of an input line card, a control module and at least one output line card, the number of which depends on the number of channels to be simultaneously processed. The functions performed by the router allow both the processing of the 18 GHz-SCM optical label at 155 Mbit/s and the routing and forwarding of the optical packets at an effective data rate of 10 Gbit/s.
The input line card consists of the filtering subsystem which operates on the combined signal and extracts the routing information carried in the SCM encoded label. The extraction process is accomplished directly in the optical domain by using a combination of an Arrayed Waveguide Grating (AWG), acting as a demultiplexer and a set of high performance Fiber Bragg Gratings (FBG) with resonance frequencies equal to the central frequencies of the incoming packets, featuring a bandwidth of 0.18 nm [6, 7]. The configuration of the filtering subsystem reflects the optical carrier where data packets are conveyed in with 22 dB SCM label suppression whereas the optical label with 50 dB carrier suppression is directed to the control module subsystem, where it is processed separately. The buffer section delays the optical packet under processing by a time required to process the label, thus achieving a synchronized attachment of the new label to the corresponding packet in the label rewriting stage.
The control module is based on a Field Programmable Gate Array (FPGA), which detects the incoming label, compares the information contained in the 32 bits label with its internal routing table and generates a new label with the routing information for the next router. The control module defines the new outgoing wavelength for each channel by driving the tunable lasers (TL1 – TL2) in order to set the output port of the router for each incoming data packet.
The output line card consists of the multistage 2R regeneration and label re-writing stages. The multistage regeneration module includes a wavelength conversion stage, based on Cross Gain Modulation (XGM), which converts the input signal to an internal fixed service wavelength to allow the assignment of any input to any output wavelength in the router. As mentioned before, the tunable lasers driven by the control module set the output wavelength for each packet by feeding the second wavelength conversion stage based on Cross Phase Modulation (XPM).
The control module incorporates electrical mixers which produce a double sideband modulation of the subcarrier at 18 GHz with the new baseband label at 155 Mb/s. The rewriting process is accomplished by combining the new label at 18 GHz with the regenerated signal using a FBG array (FBGA) and the circulator. The FBGA allows the transmission of the SCM label and rejects the optical carrier whereas in counter-propagation mode, the FBGA rejects the signal coming from the regeneration stages and combines it with the transmitted SCM label. Each output line card processes a single incoming wavelength channel at a time, thus upgrading the router is as easy as adding a new output line card for each wavelength channel supported by the network. The AWG router (AWGR), shared by all output cards, performs wavelength routing, where the output wavelength determines the node output port.
3. Modelling results and dimensioning
The core system was simulated using Virtual Photonics Inc. software. Figure 2 shows the system schematic used to simulate the node, both XGM and XPM wavelength conversion stages as well as demultiplexers and passive routing devices were included in the simulation. The virtual system tries thus to emulate the behaviour of a complete core router, in accordance with the laboratory measurements performed earlier. Note that only the packet payload path was simulated since we were interested in assessing the cumulative effects on the packets at 10 Gbit/s caused by the cascade of several two 2R regenerators. The aforementioned devices have a direct impact on the maximum number of hops which can be supported by the network structure. Both 2R regenerators based on SOAs include the dependence of the gain caused by saturation effects and time dependent phase change due to the gain index coupling.
The 10 Gbit/s payload is transmitted at 1552.5 nm through the AWG to the SOA based XGM wavelength converter stage, where it is translated to the service wavelength at 1553.3 nm. The SOA injection current was set at 1A and the carrier density at transparent point was 1.2×1024 1/m3, A band-pass filter with central frequency at the service wavelength filters out everything but the converted signal, while the attenuator boosts the signal to the power level suitable for the second wavelength converter, which uses XPM based on an arrangement of two SOAs in a Mach-Zehnder interferometer configuration. The injection current for the upper SOA is 0.450 A whereas the lower one was driven with 0.300 A. Both amplifiers share the same carrier density at 1.4×1024 1/m3. The AWG demultiplexer and AWG router only impose insertion losses to the signal.
The examined system was simulated in a network structure with up to 64 hops and Fig. 3 shows the Bit Error Rate (BER) for every single hop within the examined network. The multistage conversion router was set-up to emulate the laboratory measured parameters as close as possible. The BER curves show a power penalty at 16 hops of 1.3 dB with BER=10-12 when compared with the back-to-back curve. For network structures with more than 16 and less than 64 hops, the penalties become higher but no error floor was observed in the executed simulations. With more than 64 hops, the accumulated timing jitter imposes a strong penalty with the observable error floor at 10-9.
4. Experimental performance of the node
The all-optical router was first characterized in order to evaluate its performance. Experimental measurements for 10 Gbit/s NRZ modulated payload including optical spectra and jitter contributions are shown in Fig. 4, including the eye diagrams shown as insets. The input data stream is modulated on a 1550.9 nm carrier and then injected into the intra-node regeneration stages, as depicted in Fig. 4(a). The signal has a quality factor of 13. Figure 4(b) shows the output of the XGM stage based wavelength converter Kamelian, which was fed by a CW source operating at 1534.186 nm. The bandwidth limitation and the noise behaviour of the SOA degrade the signal in such a way that it has the resulting quality factor of 8. The output of the second regenerator, based on a Heinrich-Hertz-Institute (HHI) SOA-XPM, is depicted in Fig. 4(c). In this case, the output wavelength centred at 1550.9 nm is the same as the input wavelength. Different combinations of input/output wavelengths were tested and found to lead to the same results.
The XPM stage performs rapid conversion with high extinction ratio for the baseband delivering a signal with a quality factor of 12.5 at the output of the router. The timing jitter was measured at the node input and after the regenerating stages at several received signal powers. As expected, the total jitter contribution is slightly higher after each stage with higher values for lower received powers, e.g. the jitter of the incoming signal at a received power of -12 dBm increased from 5.3 ps to 6.4 ps after the first regenerator and 7 ps after the cascade of two 2R regenerators.
5. Multihop packet transport experiment
This section describes the experimental examination of a multihop operation of the all-optical router with label swapping capabilities in accordance with the description included in section 2. The experiment allowed us to test the signal path corresponding to transmission of the 10 Gbit/s packets. The setup consists of a single input line card, one output line card, the control module and an additional AWGR at the input of the node, which is used to close the loop and route the feed backed packets with different wavelengths to the node input port. Figure 5 shows the experimental setup.
The generated signal is transmitting through the WGR 2 instead to the AWGR 1 for synchronization purposes since data packets have to reach the node input 2.5 μs after they had been transmitted. Note that we do not use an optical switch to assemble the looped structure but it is rather achieved by the means of an accurate synchronization of the active/inactive states for the tunable lasers (TL 1, TL 2), feeding the packet generator and the XPM stage in the output line card.
The generator produces two types of packets with 231-1 PRBS pattern. Packet_1 is assigned wavelength label_1 transmitted at 1550.9 nm and it is used to fill the data loop. Packet_2 with wavelength label_2 transmitted at 1559.3 nm is used to keep the system in synchronized state. The duration of both packets is set at 2.5 μs which is equivalent to single loop transition time. In the experiment description, we use the following nomenclature: one cycle is defined as a single loop transition time, thus the system configured for 1 hop operation starts by sending packet_1, which traverses from input port 2 to output port 5 in the AWGR 2, thus packet_1 enters the loop completing the first cycle.
In the second cycle, packet_1 reaches the output line card, which routes the packet_1 from input port 1 to output port 5. In this way, the packet starts the second cycle inside the router module while simultaneously the packet generator transmits packet_2 from input port 2 to output port 10, thus packet_2 is dropped. In the third cycle, packet_1 that just finished two loops inside of the node, which is effectively equal to a single network hop, is routed from input port 1 to output port 6. Thus, packet_1 is extracted from the loop to be analyzed. Once the packet is extracted, a new packet_1 is produced by the generator and the above described steps are repeated over and over again. The same procedure is used to test the packet transport in 3, 7 and 16 hop systems. The only observable difference lies in the number of times that packet_2 is generated between two consecutive packets_1, because of the fact that the generator will not send a new packet_1 until the previous one completes the pre-defined number of hops. Figure 6 shows the measured stream of packets (left) and its respective timing synchronization between the two types of packets used for each configuration. Additionally, the wavelength assignment for the experiment is also depicted (right). Boxes with diagonal lines do not represent two consecutive packet_1 but rather the wavelength conversion process for that packet inside the router. Note also that packet_1 is under processed while packet_2 is being transmitted.
Figure 7 shows the eye diagrams of the payload, the observable degradation as the number of hops increases is due to the accumulated jitter generated by the two 2R regenerators in combination with the effect of the low extintion ratio and bandwidth limitation imposed by the XGM wavelength conversion stage at each pass through the router . For instance, the quality factor of the incoming signal at a received power of -10 dBm is degraded from Q = 6.7 to Q = 6.5 after the first hop, Q = 6.1 after 3 hops, Q = 5.5 after 7 hops and finally Q = 5.1 after 16 hops.
The observed system penalties impact the BER performance of the examined network. The filled points in Fig. 8 depict the experimental BER curves for a single router pass and 1, 3, 7 and 16 hops. The estimation on the system penalty induced by 16 hops was 1 dB with BER = 10-12 and 2 dB with BER = 10-9. The figure also shows in solid lines the theoretical BER performance results obtained from the simulations. As expected, the behavior is similar to the experimental results validating the theoretical predictions for network dimensioning, depicted first in Fig. 3.
We proposed and demonstrated an all-optical label swapping router capable of processing two parallel wavelength channels with data packets at 10 Gbit/s and SCM encoded labels at 155 Mbit/s. The router performs label extraction, processing and re-writing functions and incorporates two 2R regenerators: the first one is based on XGM and translates the incoming packets at 10 Gbit/s to a fixed service wavelength and the second one is based on XPM and converts the packet at the service wavelength to the final packet wavelength. The combination of the two regenerators allows the assignment of any input wavelength channel into any output wavelength channel. Experimental measurements emulating the packets traversing the router up to 16 hops show a penalty of 1 dB with BER = 10-12 whereas in the simulation the measured penalty for the same number of hops was 1.3 dB. The error floor appears for systems with more than 64 hops. Experimental characterization and subsequent validation through simulations demonstrate the feasibility of the system to be used in future IP over WDM networks.
The authors wish to acknowledge the E.U. funded project IST - LABELS, The work of G. Puerto was supported by the Government of Valencia (Generalitat Valenciana).
References and links
1. M. Y. Jeon, Z. Pan, J. Cao, Y. Bansal, J. Taylor, Z. Wang, V. Akella, K. Okamoto, S. Kamei, J. Pan, and S. J. Ben Yoo, “Demonstration of all-optical packet switching routers with optical label swapping and 2R regeneration for scalable optical label switching network operation,” J. Lightwave Technol. 21, 2723–2733 (2003). [CrossRef]
2. Z. Zhu, V. J. Hernandez, M. Y. Jeon, J. Cao, Z. Pan, and S. J. Ben Yoo, “RF photonics signal processing in subcarrier multiplexed optical-label switching communication systems,” J. Lightwave Technol. 21, 3155–3166 (2003). [CrossRef]
3. T. Koonen Sulur, I. T. Monroy, J. Jennen, and H. de Waardt, “Optical labeling of packets in IP-over-WDM networks,” Proc ECOC (2002), Paper 5.5.2.
4. S. J. Ben Yoo, H. J. Lee, Z. Pan, J. Cao, Y. Zhang, K. Okamoto, and S. Kamei “Rapidly switching all-optical packet routing system with optical-label swapping incorporating tunable wavelength conversion and a uniform-loss cyclic frequency AWGR,” IEEE Photon. Technol. Lett. 14, 1211–1213 (2002). [CrossRef]
5. Z. Zhu, M. Funabashi, Z. Pan, L. Paraschis, and S. J. Ben Yoo, “10000-hop cascaded in-line all-optical 3R Regeneration to achieve 1250000 km 10-Gb/s transmission,” IEEE Photon. Technol. Lett. 18, 718–720 (2006). [CrossRef]
6. M. Popov, A. Martinez, J. Capmany, D. Pastor, P. Y Fonjallaz, and B. Ortega, “Fiber-Bragg-grating-based device for payload and label separation in highly packed subcarrier-multiplexed optical label swapping,” IEEE Photon. Technol. Lett. 17, 2445–2447 (2005) [CrossRef]
7. H. J. Lee, S. J. Ben Yoo, V. K. Tsui, and S. K. H. Fong ,”A simple all-optical label detection and swapping technique incorporating a Fiber Bragg Grating filter,” IEEE Photon. Technol. Lett. 13, 635–637 (2001). [CrossRef]