We present results from the first demonstration of a fully integrated SDN-controlled bandwidth-flexible and programmable SDM optical network utilizing sliceable self-homodyne spatial superchannels to support dynamic bandwidth and QoT provisioning, infrastructure slicing and isolation. Results show that SDN is a suitable control plane solution for the high-capacity flexible SDM network. It is able to provision end-to-end bandwidth and QoT requests according to user requirements, considering the unique characteristics of the underlying SDM infrastructure.
© 2014 Optical Society of America
The dramatic growth of Internet traffic is widely recognized by operators and market analysts alike. This trend is likely to continue due to the proliferation of fiber to the premises (FTTP) and other means of high bandwidth access, as well as to the emergence of disruptive, bandwidth hungry applications such as video-on-demand, high-definition TV, Cloud and Grid Computing. Thus, in the next 10 to 15 years, the Internet is expected to undergo major transitions with respect to technologies, services and, especially, its size. As shown in Fig. 1(a), current rates indicate a growth of more than 30% every year  that if sustained will lead to growth of the Internet of up to 50 times within the next 15 years.
Meanwhile, Space Division Multiplexing (SDM) has been shown to provide a dramatic increase in transmission capacity [3,4]. As shown in Fig. 1(b), in spite of the early stages in the development of SDM technology, there have been several demonstrations of record transmission capacities with similar gains (10 times transmission capacity increase in just 2-3 years) with respect to WDM as those shown by early WDM technology over TDM in the 1990s. Additional benefits of SDM technology, such as the possibility to support spatial superchannels  and Self-Homodyne Detection (SHD)  have been shown to relax the receiver complexity and laser linewidth requirements by exploiting the highly correlated properties of different cores in a Multi-core Fiber (MCF). Optical nodes that support switching of self-homodyne channels have also been recently demonstrated . Other work has explored the benefits of SDM to support new networking functionalities beyond the straightforward increase in capacity, i.e. to provide additional flexibility in bandwidth provisioning .
However, the increased capacity, flexibility and complexity of optical SDM networks represent new challenges for the network control and management. For instance, an SDM network control carrying self-homodyne spatial superchannels is required to simultaneously calculate end-to-end paths for spatial subcarriers and pilot tones (PT), ensuring that the phase correlation between the subcarriers and PT is not lost. This requires the control plane to distinguish between links with multiple fibers, which characteristics are not necessarily similar, and multiple cores in a multi-core fiber (MCF), which would exhibit very similar characteristics. Furthermore, if dynamic node architectures are used such as Architecture on Demand (AoD) nodes  the control plane should be able to dynamically synthesize the node architecture according to the requirements of traffic. Also, a suitable control plane for SDM should support virtualization of network resources, in order to create multiple isolated slices of the optical infrastructure, which can be allocated, controlled or even programmed by different network services, facilitating emerging requirements for efficient and dynamic infrastructure sharing. Therefore, to fully exploit SDM networking, it is necessary to develop novel approaches in network control and management, which are able to support the increased complexity of the transport layer and facilitate new network functionalities enabled by the additional spatial dimension. Here, we propose to utilize Software Defined Networking (SDN) for control and management of novel bandwidth-flexible and programmable SDM networks based on MCF links, AoD nodes and carrying sliceable self-homodyne spatial superchannels. The SDN control enables service level programmability of network and transport functions by abstracting the technological details of the SDM infrastructure, making it sliceable, directly accessible and controllable by network services. This is achieved by exploiting intrinsic characteristics of SDN, such as the separation of data plane and control plane, the availability of a centralized controller and view of the network, the use of open interfaces between the devices in the control plane and the agents in the data plane, and the possibility to program the network by external applications.
In this paper, we present results from the first demonstration of SDN-enabled control plane, fully controlling the node architecture configuration and bandwidth provisioning, over an SDM network that consists of three Architecture on Demand nodes, linked by two MCFs with 19 and 7 cores, carrying sliceable self-homodyne spatial superchannels. Such superchannels support multiple bit rates by varying the number of aggregated spatial subcarriers and their modulation format, which can be selected between quaternary phase shift keying (QPSK) and 16 quadrature amplitude modulation QAM. Idle subcarriers, which are not used by existing spatial superchannels, are allocated to new superchannels for transmission to other destinations. Therefore, it is possible to support superchannels with up to 18 spatial subcarriers or partition the subcarriers into smaller spatial superchannels, i.e. superchannel slicing. AoD nodes dynamically implement node architectures tailored to traffic requirements, thereby using the available technology in a flexible manner. In the control plane, a novel SDM Flow Mapper and a multi-dimensional bandwidth slicing service are shown to successfully map network application requirements, i.e. bandwidth and Quality of Transport (QoT), to the self-homodyne spatial superchannel slices, the required transmission technology (e.g. MCF or SMFs) according to channel characteristics, and the switching technology available in the AoD nodes.
2. Experimental SDM network data plane setup
As shown in Fig. 2, the experimental setup is comprised of three programmable nodes and three transmitters. Nodes 1-3 implement the AoD concept  and consist of an optical backplane based on 3D-MEMS  that interconnects MCF/SMF fiber inputs, modules (e.g. Spectrum Selective Switch (SSS), amplification stages, etc.) and MCF/SMF fiber outputs, as shown in Fig. 3(a). AoD makes it possible to dynamically provide synthetic node architectures tailored to traffic requirements. Different synthetic architectures are implemented by interconnecting fiber inputs, the modules available in the AoD node, and fiber outputs, using cross-connections in the optical backplane. One important advantage of AoD nodes is that modules are used only when required. For instance, in case all channels from an input core need to be switched to the same output a single cross-connection is required to switch potentially large traffic volumes, which improves the scalability of the system. If the same channels require amplification, an erbium-doped fiber amplifier (EDFA) is connected between the node’s input and output, avoiding the requirement for additional devices such as splitters, (de)multiplexer, etc.
Nodes 1 and 2 are connected by a 10-km 19-core fiber (MCF-1) with an average intercore crosstalk of −37.2 dB. Nodes 2 and 3 are connected by a 30-km 7-core fiber (MCF-2) with an average intercore crosstalk of −54 dB. Insets A and B of Fig. 2 show the MCF facets, which have a core pitch of 35µm and 56µm for MCF-1 and MCF-2 respectively. SDM MUX/DEMUX devices based on laser-inscribed 3-D waveguide are used to interconnect the MCFs with the SMF-compatible optical backplane, as illustrated in Fig. 3(a). The average total loss of the MCF including SDM MUX/DEMUX was measured to be 5.2 dB and 9.8 dB for MCF-1 and MCF-2, respectively. Also, four independent Single Mode Fiber (SMF) links directly interconnect Nodes 1 and 3. Tx-1 generates sliceable self-homodyne spatial superchannels over 14 wavelengths. Each spatial superchannel consists of a programmable number of spatial subcarriers (from 1 to 18) with 64-Gb/s single-polarization QPSK or 128-Gb/s 16QAM modulation and a pilot tone, all at the same wavelength but using different cores, as shown in Fig. 3(b). PTs are transmitted together with the modulated signals and are used at the receiver side for coherent detection, obviating the need for a local oscillator (LO). As PTs and modulated carriers are generated from the same source, the phase noise is cancelled out when the signals are mixed at the receiver, provided that the path lengths are the same. Therefore, it is possible to relax the laser linewidth requirements, thereby making it possible to use low-cost distributed feedback (DFB) lasers. Therefore, savings by using self-homodyne transmission are twofold, firstly, by making it possible to use low-cost DFB lasers in the transmitter and, secondly, by eliminating the need for a LO in the receiver.
Tx-1 is connected to Node-1 and is used by the SDN control plane for provisioning flexible-bandwidth channels towards Node-2 or Node-3. Tx-2, connected to Node-1, generates Binary Phase Shift Keying (BPSK)-modulated 40-Gb/s channels. Tx-3, connected to Node-2, generates seven self-homodyne channels with 20-Gb/s QPSK modulation. PTs generated at Tx-1 and Tx-3 are labeled PT1 and PT3 respectively, and are required to demodulate the QPSK and 16QAM signals, as they are generated from DFBs with large linewidths. To retain their correlation, the modulated spatial subcarriers and their corresponding PT are generated, added to the nodes, transmitted (through the same MCFs), and dropped together. The architecture implemented by AoD is calculated and implemented automatically by the SDN control plane by configuring cross-connections in the optical backplane.
3. SDN over SDM network features and benefits
The SDM network described in the previous section enables efficient infrastructure utilization and high scalability through the introduction flexibility and programmability. For instance, the transmission capacity of the sliceable self-homodyne spatial transmitter can be fully utilized, as variable-bandwidth superchannels can be setup and idle spatial subcarriers can be used to transmit to other destinations. Similarly, hardware resources are efficiently used in the AoD nodes, as hardware modules are used only when and where required. It is also possible to split SDM network resources (e.g. using different cores or spectrum) and allocate separate partitions to users/applications, i.e. readily supporting infrastructure virtualization. However, in order to take full advantage of these data plane features, it is necessary to implement a control plane that can seamlessly consider their advantages and restrictions, and utilize them in an optimum manner.
SDN provides four key features that make it suitable for the SDM network data plane described in the previous section:
3.1. Separation of data plane and control plane
This allows the control plane to be responsible for slicing and allocating the SDM infrastructure, as well as routing optical flows through the SDM network. Once the SDN control plane has determined the required configuration, it is implemented on the SDM network data plane. This facilitates the constant and somewhat independent evolution of the data plane and the control plane. For instance, this is a key characteristic for supporting AoD nodes with varied capabilities (i.e. different modules deployed on each node). It is also important to support the evolution of AoD nodes, through the introduction of new or enhanced functionalities; and the SDM networking infrastructure, through the deployment of new SDM links (e.g. multiple SMF, 7-core, 19-core, etc.) or transmitters with different characteristics (e.g. transmitters with sliceable space, sliceable spectrum, or self-homodyne operation). Without this characteristic it would be necessary to reconfigure the optical nodes and transceivers every time the routing, resource allocation or virtualization policy had to be modified, which would potentially require considerable effort and investment.
3.2. A centralized controller and view of the network
This makes it possible to optimize the use of SDM network resources end-to-end, taking into consideration the characteristics of the infrastructure on each area of the network. For instance, it is possible to maintain a general routing policy where channels that do not require MCFs are routed preferably through SMFs. In this manner, MCF resources are kept available for services that benefit from MCF characteristics, such as self-homodyne transmission. A centralized view of the network is also useful for optimizing the use of modules in AoD nodes, as different nodes may provide different functionalities.
3.3. Open interfaces between the devices in the control plane and the agents in the data plane
This ensures compatibility and upgradeability of the SDM network. In conjunction with the availability of drivers for network components, this characteristic is key to achieving plug-and-play operation of the SDM network. For instance, if a new element or device is deployed, it can be integrated and used with minimum effort or manual intervention.
3.4. Programmability of the network by external applications
This is enabled by abstraction of the underlying infrastructure by the SDN controller. Once the underlying SDM infrastructure has been abstracted, it can be offered as a service to higher layers, including external applications. This enables innovation, fast development and deployment of new services, and a rapid evolution of the network to cope with new requirements from users and applications.
4. Experimental SDN control plane setup
The proposed control plane architecture includes an SDN controller and a multi-dimensional network slicing service (i.e. external application), as shown in Fig. 2. The information required by the multi-dimensional slicing service to operate is generated by the process of abstraction of the underlying network infrastructure. The Open Flow (OF) interface in the controller and the OF Agent in the device utilize an extended OF protocol to abstract the transmitter states (Tx-1, Tx-2 and Tx-3), the AoD nodes’ modules and features, network topology and connectivity (SMFs and MCFs). Note that communications between the separate control and data planes are performed entirely by open interfaces. The abstraction method handles MCFs as a single entity with multiple spatial channels, which are expected to deliver very similar characteristics for transmission. This information is used by the multi-dimensional slicing service for routing self-homodyne spatial superchannels, as they need to maintain high correlation between modulated spatial subcarriers and PTs in order to cancel out their inherent phase noise at the coherent receiver. SMFs are abstracted as multiple entities with a single spatial channel each. This means they are not guaranteed to provide very similar characteristics suitable for self-homodyne spatial superchannel transmission.
We propose a new multi-dimensional Optical Flow, as shown in Fig. 2. It supports the abstracted representation of MCFs and SMFs, and signals with various center wavelength, bandwidth and modulation format. As such, this multi-dimensional optical flow is suitable for representing spatial superchannels, as well as spectral superchannels with variable bandwidths, or a combination of spatial and spectral superchannels. We have extended the switch-features message of standard OF protocol for enabling the controller to discover underlying physical layer devices, their capability and their connectivity and store them in the Topology Database. This information is exposed through the Application Northbound Interface as an abstracted topology, comprising a set of generic nodes and links. It provides a centralized view of the network to external applications (e.g. the multi-dimensional slicing service), where each node is characterized by its switching and programming capability and each link by its supported bandwidth granularity and QoT.
The multi-dimensional slicing service is an external network application based on a novel multi-criteria heuristic algorithm that has been designed and developed for this testbed. As shown in Fig. 4, this application deals with requests that consist of source, destination, bit rate and QoT (BER) requirements. It calculates a route between the given source and destination nodes considering the availability of resources in the SDM network. The first step is to decide between self-homodyne superchannel transmission through MCF or single channel transmission through SMF, depending on the required bit rate. Self-homodyne spatial superchannels over MCF links are used to provide high data rates, e.g. 512 Gb/s and 384 Gb/s, whereas SMF links are used for lower bit rates, e.g. single carrier 40 Gb/s. Next, a set of routes between the source and destination nodes are evaluated to determine if they can provide the required QoT and with which modulation format. Based on this, the most efficient modulation format that can provide the required QoT is chosen for transmission. The bit rate and modulation format are used to calculate the number of spatial subcarriers and bandwidth required. This information, together with the selected route, is passed to the multi-dimensional resource allocation engine. It assigns fibers, cores, and bandwidth in the SDM network data plane. This algorithm allocates 25-GHz bandwidth for the PTs of spatial superchannels, as they do not require wide bandwidth for transmission. If the request is successfully allocated, the information is passed to the technology mapper in the SDN controller. It translates the algorithm output into a set of cross-connections in the AoD nodes, to compose the required node architecture, and the configuration of the node components (e.g. waveshapers), to provision the required bandwidth. The SDN controller configures the physical layer resources, i.e. transmitters and AoD nodes, according to the output of the Flow Mapper using the extended OF Flow-Mod messages. As shown in Fig. 4, if there are no resources available in the SDM network the request is blocked.
To test the proposed SDN-controlled SDM network, we designed and implemented a network application that generates requests, simulating users/applications with different bandwidth and QoT (BER) requirements and different source and destination pair. Figure 5 shows the channels switched through the network. All self- homodyne spatial superchannels are switched through MCFs, as they require to maintain high correlation between the modulated signals and their corresponding PTs in order to successfully implement SHD at the receiver. On the other hand, 40Gb/s BPSK channels are switched through SMFs, as they do not require SHD. We demonstrate flexible bandwidth provisioning (384 Gb/s, 512 Gb/s and 512 Gb/s) with different QoT requirements (BER<1e-5, BER<1e-5 and BER<2e-3 respectively) using three slices of the self-homodyne spatial superchannel transmitter at 1558.98 nm, as shown in Figs. 5(a)-5(c). The SDM data plane supports two mechanisms for varying bit rates, namely, by varying the number of spatial subcarriers and by using different modulation format in each self-homodyne superchannel.
Spatial superchannel slices are successfully provisioned by the SDN control plane, which configures Tx-1 (e.g. to select 16QAM for Slice-3), calculates and configures the signal and PT paths through the SDM network (e.g. fiber, core and wavelength selection, AoD cross-connections and SSS pass-bands). The spectra of the 64 Gb/s single polarization (SP)-QPSK and 128 SP-16QAM signals generated in Tx-1 are shown in Figs. 6(a) and 6(b) respectively. Depending on the choice of modulation format, these signals are selected, replicated and output onto each of the cores of MCF1. For instance, the output of Tx-1 for MCF-1 core 18 consists of five spatial subcarriers at 64 Gb/s and six spatial subcarriers at 128 Gb/s. These signals are all part of different self-homodyne spatial superchannels but have been WDM multiplexed onto a single core, as shown in Fig. 6 (c). In order to reduce crosstalk on the shared WDM PTs, the algorithm routes PT1 through core 19, which is one of the external cores of MCF-1. Similarly, at Node-2, one copy of PT1 is dropped for SHD of the dropped channels, while another is transmitted over MCF-2 core 7. The 20 Gb/s SP-QPSK signals from Tx-3 input to Node-2 [Fig. 6(d)] are WDM multiplexed with 64 Gb/s QPSK subcarriers coming from Tx-1 on core 5 of MCF1 [Fig. 6(e)], and transmitter over core 5 of MCF2 [Fig. 6(f)]. PT3 is transmitted over MCF-2 core 6, after WDM multiplexing with 64-Gb/s signals from Tx-1, as shown in Figs. 6(g)-6(i). WDM multiplexing in Node-2 is implemented using spectrum selective switching devices based on liquid crystal on silicon (LCoS) , which enable support for flexible spectrum switching . Therefore, the pass band used for WDM multiplexing at Node-2 is tailored to the requirements of each signal. Thus, 64 Gb/s SP-QPSK self-homodyne subcarriers from Tx-1, 20 Gb/s SP-QPSK from Tx-3 and pilot tones from Tx-3 (PT3) are switched using bandwidths of 100 GHz, 50 GHz and 25 GHz respectively. This demonstrates that the testbed is also capable of supporting flexible allocation of spectral resources, e.g. as required for elastic optical networking  .
The signaling exchange between the OF Agent, the SDN Controller and the multi-dimensional slicing service, used to setup Slice-1 [Fig. 5(a)] is shown in Fig. 7. It shows the OF topology discovery features messages followed by the bandwidth request sent to the slicing service solver via the controller. The HTTP web service request shows the ASCII code with the XML details sent to the slicing service i.e. 384 Gb/s with BER<1e-5 from Node-1 to Node-3. Once the slicing service finds the best solution, the flow mapper issues the corresponding Flow Mod messages. This is verified by receiving the port status update from the Agent. Inset A in Fig. 7 plots the average path setup times for Slice-1. The slicing service and the device configuration times contribute most of the setup delay, due to the complexity associated with the request and the implementation of the OF abstraction using the device’s proprietary programming interface. These results demonstrate that an OF-based SDN solution provides a simple and reliable interface to abstract the SDM network, so intelligent network services can run seamlessly over the controller.
Figure 6(c) shows the signals on MCF-1 core 18, after provisioning Slice 3. SHD  followed by off-line analysis was used to measure the BER of the spatial superchannels. Results are presented in Fig. 8(a). The measured end-to-end BER for 16QAM and QPSK spatial subcarriers was below 2e-3 and 6e-6 respectively. Figure 8(a) also shows the QoT threshold (BER = 1e-4) used by the multi-dimensional slicing service to select between QPSK or 16QAM modulation formats. 40 Gb/s BPSK was measured error-free. Constellations of the signals used in the experiment are shown in Fig. 8(b).
We have presented the first SDN over SDM multi-granular switching network based on two MCFs, with 7 and 19 cores, and three programmable AoD nodes, supporting sliceable and bandwidth-variable self-homodyne spatial superchannels. An external multi-dimensional slicing service utilizes abstracted information about the underlying SDM infrastructure to setup services with multiple bit rates and QoT according to user requirements using the SDN control. We demonstrated SDN-controlled automatic bandwidth and QoT provisioning over the SDM infrastructure, AoD node configuration, and self-homodyne spatial superchannel slicing, routing and switching according to user requirements and with good end-to-end performance.
This work is supported by the EPSRC grant EP/ I01196X: The Photonics Hyperhighway, and the EC FP7, grant no. 317999, IDEALIST. The authors are grateful to Furukawa Electric from Japan for providing the 7-core MCF and to Kylia for the 16QAM emulator.
References and links
1. White paper, “Cisco visual networking index: forecast and methodology, 2010-2015
2. R. Essiambre and R. W. Tkach, “Capacity trends and limits of optical communication networks,” Proc. IEEE 100(5), 1035–1055 (2012). [CrossRef]
3. J. Sakaguchi, B. J. Puttnam, W. Klaus, Y. Awaji, N. Wada, A. Kanno, T. Kawanishi, K. Imamura, H. Inaba, K. Mukasa, R. Sugizaki, T. Kobayashi, and M. Watanabe, “19-core ﬁber transmission of 19x100x172-Gb/s SDM-WDM-PDM-QPSK signals at 305Tb/s,” in Optical Fiber Communication Conference. Optical Society of America, 2012, p. PDP5C.1.
4. H. Takara, A. Sano, T. Kobayashi, H. Kubota, H. Kawakami, A. Matsuura, Y. Miyamoto, Y. Abe, H. Ono, K. Shikama, Y. Goto, K. Tsujikawa, Y. Sasaki, I. Ishida, K. Takenaga, S. Matsuo, K. Saitoh, M. Koshiba, and T. Morioka, “1.01-Pb/s (12 SDM/222 WDM/456 Gb/s) crosstalk-managed transmission with 91.4-b/s/Hz aggregate spectral efficiency,” in European Conference and Exhibition on Optical Communication, Optical Society of America, 2012, p. Th.3.C.1. [CrossRef]
5. L.E. Nelson, M.D. Feuer, K. Abedin, X. Zhou, T.F. Taunay, J.M. Fini, B. Zhu, R. Isaac, R. Harel, G. Cohen, and D.M. Marom, “Spatial superchannel routing in a two-span ROADM system for space division multiplexing,” J. Lightwave Technol., early access articles, (2013)
6. B. J. Puttnam, J. M. D. Mendinueta, J. Sakaguchi, R. S. Luis, W. Klaus, Y. Awaji, N. Wada, A. Kanno, and T. Kawanishi, “210Tb/s self-homodyne PDM-WDM-SDM transmission with DFB lasers in a 19-core fiber,” in Photonics Society Summer Topical Meeting Series, 2013, p. TuC1.2.
7. J. Sakaguchi, W. Klaus, B. J. Puttnam, T. Miyazawa, Y. Awaji, J. M. Delgado Mendinueta, R. S. Luis, and N. Wada, “SDM-WDM hybrid reconfigurable add-drop nodes for self-homodyne photonic networks,” in Photonics Society Summer Topical Meeting Series, 2013, p. WC1.2.
8. N. Amaya, M. Irfan, G. Zervas, R. Nejabati, D. Simeonidou, J. Sakaguchi, W. Klaus, B. J. Puttnam, T. Miyazawa, Y. Awaji, N. Wada, and I. Henning, “First fully-elastic multi-granular network with space/frequency/time switching using multi-core ﬁbres and programmable optical nodes,” in European Conference and Exhibition on Optical Communication. Optical Society of America, 2012, p. Th.3.D.3.
9. N. Amaya, G. S. Zervas, B. R. Rofoee, M. Irfan, Y. Qin, and D. Simeonidou, “Field trial of a 1.5 Tb/s adaptive and gridless OXC supporting elastic 1000-fold bandwidth granularity,” in Optical Communication (ECOC), 2011 37th European Conference and Exhibition on, Sept. 2011. [CrossRef]
10. M. C. Wu, O. Solgaard, and J. E. Ford, “Optical MEMS for lightwave communication,” J. Lightwave Technol. 24(12), 4433–4454 (2006). [CrossRef]
11. G. Baxter, S. Frisken, D. Abakoumov, H. Zhou, I. Clarke, A. Bartos, and S. Poole, “Highly programmable wavelength selective switch based on liquid crystal on silicon switching elements,” in Optical Fiber Communication Conference, 2006 and the 2006 National Fiber Optic Engineers Conference. OFC 2006, March 2006. [CrossRef]
12. N. Amaya, M. Irfan, G. Zervas, K. Banias, M. Garrich, I. Henning, D. Simeonidou, Y. R. Zhou, A. Lord, K. Smith, V. J. F. Rancano, S. Liu, P. Petropoulos, and D. J. Richardson, “Gridless optical networking ﬁeld trial: flexible spectrum switching, defragmentation and transport of 10G/40G/100G/555G over 620-km ﬁeld ﬁber,” in 37th European Conference and Exhibition on Optical Communication (ECOC), 2011, Sept. 2011. [CrossRef]
13. O. Gerstel, M. Jinno, A. Lord, and S. J. B. Yoo, “Elastic optical networking: a new dawn for the optical layer?” IEEE Commun. Mag. 50(2), S12–S20 (2012). [CrossRef]
14. M. Jinno, H. Takara, B. Kozicki, Y. Tsukishima, Y. Sone, and S. Matsuoka, “Spectrum-eﬃcient and scalable elastic optical path network: architecture, beneﬁts, and enabling technologies,” IEEE Commun. Mag. 47(11), 66–73 (2009). [CrossRef]