By utilizing the cyclic filtering function of an NxN arrayed waveguide grating (AWG), we propose and experimentally demonstrate a novel multi-function all optical packet switching (OPS) architecture by applying a periodical wavelength arrangement between the AWG in the optical routing/buffering unit and a set of wideband optical filters in the switched output ports to achieve the desired routing and buffering functions. The proposed OPS employs only one tunable wavelength converter at the input port to convert the input wavelength to a designated wavelength which reduces the number of active optical components and thus the complexity of the traffic control is simplified in the OPS. With the proposed OPS architecture, multiple optical packet switching functions, including arbitrary packet switching and buffering, first-in-first-out (FIFO) packet multiplexing, packet demultiplexing and packet add/drop multiplexing, have been successfully demonstrated.
© 2012 OSA
With the increases of both the versatile Internet services and wideband access subscribers, a goal for the optical transport and switching system designs is to meet the explosive bandwidth demands. The advanced dense wavelength division multiplexing (DWDM) technology is now capable of supporting more than 100 Tbps data within a single optical fiber in lab demonstration . Though large capacity optical fiber networks have been heavily deployed, the networks efficiency is still far behind the desperate need. One major obstacle limiting the networks efficiency is that considerable optical-electrical-optical (O/E/O) conversions still exist in current optical switching networks. To reduce the limitations from O/E/O processes, it is a laudable goal to develop all optical packet switching (OPS) technologies to overcome the electronic bottlenecks, for example: an all optical switch that can achieve up to 160 Gbps data stream has been demonstrated . Moreover, since the optical buffers play a key role in resolving packet contentions in the OPS, efficient packet buffering in the OPS is also an important issue toward all optical packet switching networks.
A popular configuration to achieve an OPS is constructed with fast tunable wavelength converters (TWC) and wavelength selective routers [3–10] for providing the optical packets with routing capability and a set of fixed or reconfigurable optical buffers  for buffering function. The major reconfigurability relies on the utilization of active components in the OPS. However, the employed active devices typically introduce more signal degradations, such as ASE accumulation in wavelength conversion, signal chirping in slow light buffers [8,9] and limited high speed switching extinction in monolithic optical switch . Meanwhile, more active devices require more control signals which increase the design complexity of the traffic controller. Therefore, it is an important issue to reduce the employment of the required active components, while still retaining a flexible switching functionality simultaneously in an OPS. In this work, by adopting the cyclic filtering function of an NxN arrayed waveguide grating (AWG), we propose a periodic wavelength arrangement between the optical buffering units and routing elements in an OPS to experimentally demonstrate a multi-function all optical packet switching architecture with reduced control complexity. Multiple optical packet switching functions, including arbitrary packet switching and buffering, first-in-first-out (FIFO) packet multiplexing, packet demultiplexing and packet add/drop multiplexing, are successfully demonstrated at 10 Gbps. Moreover, due to the nature of wavelength discrimination mechanism provided by the cyclic filtering and FSR characteristics of an NxN AWG, multiple optical packets can be buffered simultaneously in the same optical buffer without any contention in the proposed OPS.
2. Proposed architecture and operation principle
Figure 1 schematically shows the proposed OPS architecture, in which the wavelength arrangement for converting the incoming optical packets at one of the input ports is also illustrated. At each input port, a tunable wavelength converter (TWC) is used to convert each input optical packet to an appropriate wavelength, which is designated by the traffic controller according to a specifically scheduled buffering time and planned output port. The optical buffer is achieved by a piece of fiber delay line (FDL) which is connected to one of the AWG’s output ports. The length of each FDL is selected to correspond to an n-fold optical packet time, where n is an integer from 0 to (N–1). With the cyclic filtering and free spectral range (FSR) characteristics, each AWG’s output port can sift N wavelengths, uniformly spaced by one FSR. Meanwhile, the wavelengths at any two adjacent AWG’s output ports are separated by one DWDM channel spacing of the employed AWG. The converted wavelength is routed to a piece of FDL to obtain a specific buffering time. After the optical buffer array, an N-to-N optical coupler is utilized to bridge the buffered packets from each FDL to all the output ports. Then, a fixed wideband optical filter (WBOF) which passes a set of wavelengths within one FSR of the NxN AWG is employed at each output port to select the desired optical packets. Therefore, the converted wavelength can be allocated at the y-th DWDM channel in the x-th FSR of the employed AWG by the TWC. The y-th channel is routed to a specified FDL to obtain a specific delay time by the NxN AWG and the x-th FSR is sifted by the corresponding WOBF at the x-th output port. With this routing scenario, each optical packet can be encoded with a specific output port, x, along with a designated optical delay, y, after wavelength conversion by the TWC, as shown in Fig. 1.
The detailed wavelength arrangement considering all the input ports, buffering times and output ports is tabulated in Table 1 , which is derived from the periodic filtering property of the AWG and the FSR region selected by the employed WBOFs. Each unique wavelength in the table is marked as “λxy” where the subscript x denotes the corresponding FSR region, i.e., the output port, and subscript y indicates a sequential wavelength in one FSR region which corresponds to a specific buffering time depending on the packet’s input port and required buffering time. As previously described, by properly controlling the wavelength conversion in TWC, the incoming packets can be accordingly routed to the desired output ports at the designated time slot through the passive AWG and WBOFs. Such a look-up table can be easily stored in the traffic controller to determine an adequate optical wavelength of each incoming optical packet. Moreover, since the NxN AWG will not route the same wavelength to the same output port from different input ports, optical packets with the same wavelength will not appear in the same optical buffer concurrently. Hence the proposed OPS can simultaneously buffer up to N optical packets with different wavelengths in each FDL. The maximum storage capacity of the proposed OPS can support up to N(N – 1) packets at the same time for an NxN OPS.
The system performance and operation complexity of an OPS depend on the number of required tunable/active optical devices. We compared several published architectures, including: re-circulating [3,4], feedback , feed-forward , multi-channel fiber Bragg grating (MFBG)  types, and buffering decomposition algorithm , with the proposed system in terms of required tunable components counts and control complexity, as shown in Table 2 . Considering a system scaled with N input/output ports and a storage capacity of B packets, the proposed architecture requires the least amount of tunable optical components under the same system scale and storage capacity. Therefore, it results in the least traffic control complexity. Moreover, since the employed AWG and WBOFs are all passive, the proposed architecture exhibits a stable performance by receiving the least interferences from the tuning devices and provides a cost effective solution to realize the OPS.
3. Experimental setup and results
Following the proposed configuration, an experimental setup for 8 input/output OPS is depicted in Fig. 2 . The system consists of 8 TWCs, one at each input port; one 8x8 AWG with 100 GHz channel spacing; 8 FDLs for providing buffering delay times from 0 to 7 packet delays; one 8-to-8 optical coupler, acting as a bridge to connect the optical buffering unit and the output ports; and 8 WBOFs (Alnair Labs: BVF-100), one at each output port with bandwidth setting of 6.4 nm, equivalent to one FSR of the employed 8x8 AWG. The major active component, TWC, comprises a semiconductor optical amplifier based Mach-Zehnder interferometer (SOA-MZI, Alphion: ISM4) and a fast tunable laser based on sampled-grating distributed Bragg reflector (SGDBR, Intune: AltoWave 1100) laser. The wavelength conversion is achieved by counter-propagating cross phase modulation mechanism , as shown in the inset of Fig. 2. The advantage of using counter-propagating mechanism is to avoid the employment of a fast tunable optical filter, which has to be synchronous with the wavelength conversion, to reject the original optical packet and select the wavelength-converted signal after the TWC. Moreover, the embedded SOA in the SOA-MZI can also compensate for the power loss from other passive optical components, such as AWG, FDLs, optical coupler and WOBFs. The traffic controller is implemented by a field programmable gate array (FPGA, Xilinx: Virtes-II Pro) developing system. The incoming packet sequences, with a fixed 200 ns packet length, are generated by independent pulse pattern generators at 10 Gbps with (231-1) non-return-to-zero (NRZ) pseudo random bit sequence (PRBS) patterns under burst mode operation.
In the experimental demonstration, for simplicity, one 2x2 OPS with three buffer sizes is demonstrated because such a simple setup can easily be scaled up to achieve the proposed NxN OPS. For better packet discrimination, two packet formats, a1(t) with 150 ns packet length and 50 ns guard time and a2(t) with 100 ns packet length and 100 ns guard time, are applied in the experiment. The proposed OPS architecture is capable of executing multiple switching operations. In Fig. 3 , we demonstrate four switching functions, including arbitrary packet delay and switching (S1), FIFO packets multiplexing (S2), packets demultiplexing (S3), and optical add/drop multiplexing (S4). To interpret the wavelength arrangements and buffering schedules of all the optical packets, we illustrate the optical packets’ wavelengths and schedules by a three dimensional coordinate plot. In this diagram, only two FSRs, each containing four wavelengths, are displayed. To prevent from contention, we assume that the packets at a1(t) are with a higher priority. In the demonstration of arbitrary packet buffering and switching (S1), the first packets in both a1(t) and a2(t) are planned to be delayed by two packet times and then be switched to the output port b2(t) and b1(t), respectively. According to Table 1, the first packets in a1(t) and a2(t) should be converted to λ23 and λ14, respectively. On the other hand, the remaining packets in a1(t) and a2(t) need no delay and are converted to λ11, and λ22, respectively. In the FIFO multiplexing (S2), arriving optical packets at input ports a1(t) and a2(t) are multiplexed to the same output port b1(t). According to the same priority assumption and proposed wavelength assignment, the first packet in a1(t) needs no delay, therefore the wavelength is converted to λ11. Then the first packet in a2(t) is delayed by one packet time and is converted to λ13. For the second packet in a1(t), one packet time delay is needed due to the FIFO routing policy and the assumed priority. Thus its wavelength is converted to λ12. Finally, the second packet in a2(t) is processed with two packet delay times and is converted to λ14 to prevent from contention at the output. The other two demonstrations can follow the same scenarios to obtain the corresponding results, as shown in Fig. 3.
For the bit-error-rate (BER) performance evaluation, we applied a 3 μs packet due to the slow recovering limit of the employed burst mode receiver. The measured BER curves are shown in Fig. 4 . The penalty is about 1.5 dB at BER = 10−9, which is mainly from the wavelength conversion. We expect the performance will be improved if the whole wavelength converter is monolithically integrated . Since we observed an about 0.2 dB optical power fluctuation in the employed tunable laser when wavelength transits, there exists a performance instability of about 0.3 dB maximum in the BER measurements among the four wavelength converted cases. Nevertheless, a uniform performance among all the switched packets still can be guaranteed.
In this paper, by introducing the wavelength division multiplexing (WDM) concept into the wavelength routing architecture in the proposed OPS, we significantly reduce the number of the required active optical components, therefore we can simplify the traffic control with a novel periodic wavelength arrangement. Moreover, by mainly adopting passive optical components in the routing and buffering design, the proposed OPS sustains a stable transmission performance for all of the optical packets.
This work was supported by the National Science Council of R.O.C. under contract numbers NSC 99-2628-E-007-030 and NSC 100-2628-E-007-004.
References and links
1. D. Qian, M.-F. Huang, E. Ip, Y.-K. Huang, Y. Shao, J. Hu, and T. Wang, “101.7 Tb/s (370x294 Gb/s) PDM-128QAM-OFDM transmission over 3x55-km SSMF using pilot-based phase noise mitigation,” in Proceedings of Optical Fiber Commun. Conf. (OFC) (2011), PDPB5.
2. W. Wang, L. G. Rau, and D. J. Blumenthal, “160 Gb/s variable length packet/10 Gb/s-label all-optical label switching with wavelength conversion and unicast/multicast operation,” J. Lightwave Technol. 23(1), 211–218 (2005). [CrossRef]
3. Y. N. Singh, A. Kushwaha, and S. K. Bose, “Exact and approximate analytical modeling of an FLBM-based all-optical packet switch,” J. Lightwave Technol. 21(3), 719–726 (2003). [CrossRef]
4. F. S. Choa, X. Zhao, X. Yu, J. Lin, J. P. Zhang, Y. Gu, G. Ru, G. Zhang, L. Li, H. Xiang, H. Hadimioglu, and H. J. Chao, “An optical packet switch based on WDM technologies,” J. Lightwave Technol. 23(3), 994–1014 (2005). [CrossRef]
5. C. Guillemot, M. Renaud, P. Gambini, C. Janz, I. Andonovic, R. Bauknecht, B. Bostica, M. Burzio, F. Callegati, M. Casoni, D. Chiaroni, F. Clerot, S. L. Danielsen, F. Dorgeuille, A. Dupas, A. Franzen, P. B. Hansen, D. K. Hunter, A. Kloch, R. Krahenbuhl, B. Lavigne, A. Le Corre, C. Raffaelli, M. Schilling, J.-C. Simon, and L. Zucchelli, “Transparent optical packet switching: the European ACTS KEOPS project approach,” J. Lightwave Technol. 16(12), 2117–2134 (1998). [CrossRef]
6. M. C. Chia, D. K. Hunter, I. Andonovic, P. Ball, I. Wright, S. P. Ferguson, K. M. Guild, and M. J. O’Mahony, “Packet loss and delay performance of feed back and feed forward array waveguide grating based optical packet switches with WDM inputs/outputs,” J. Lightwave Technol. 19(9), 1241–1254 (2001). [CrossRef]
7. R. Srivastava, R. K. Singh, and Y. N. Singh, “Fiber-optic switch based on fiber Bragg gratings,” IEEE Photon. Technol. Lett. 20(18), 1581–1583 (2008). [CrossRef]
8. J. Yang, N. K. Fontaine, Z. Pan, A. O. Karalar, S. S. Djordjevic, C. Yang, W. Chen, S. Chu, B. E. Little, and S. J. B. Yoo, “Continuously tunable wavelength-selective buffering in optical packet switching network,” IEEE Photon. Technol. Lett. 20(12), 1030–1032 (2008). [CrossRef]
9. R. S. Tucker, P.-C. Ku, and C. J. Chang-Hasnain, “Slow-light optical buffer: capabilities and fundamental limitations,” J. Lightwave Technol. 23(12), 4046–4066 (2005). [CrossRef]
10. H. Yang and S. J. B. Yoo, “All-optical variable buffering strategies and switch fabric architectures for future all-optical data router,” J. Lightwave Technol. 23(10), 3321–3330 (2005). [CrossRef]
11. K.-M. Feng, C.-Y. Wu, D.-H. Hsueh, C.-S. Ku, C.-P. Chang, H.-Y. Lin, J. Cheng, and J. Chen, “Demonstration of an optical FIFO multiplexer,” in Proceeding of Optical Fiber Comm. Conf. (OFC) (2008), OMN5.
12. V. Lal, M. L. Masanovic, J. A. Summers, G. Fish, and D. J. Blumenthal, “Monolithic wavelength converters for high-speed packet-switched optical network,” IEEE J. Sel. Top. Quantum Electron. 13(1), 49–57 (2007). [CrossRef]