Open Access
Add to BibSonomyAbstract
We present an open design of a filterless add/drop reconfigurable optical add/drop multiplexer module with a NETCONF northbound interface. Compared to commercial offerings, the device design is openly documented and available for detailed inspection. Performance is evaluated with up to seven adjacent high-speed 100 Gbps signals on a 50 GHz frequency grid.
© 2020 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. INTRODUCTION
Coherent detection [1] brought a significant improvement to optical networks, in terms of both extending the reach and enabling architectural changes. This paper leverages a fundamental property of coherent optical detection where the receivers “tune-in” to a desired frequency, essentially ignoring all other channels but the desired one [2,3]. This selectivity can be utilized to provide cost-effective add/drop modules for a reconfigurable optical add/drop multiplexer (ROADM). We have designed such a ROADM, which we investigate further in this paper. Moreover, we have made the design open, with the necessary northbound interfaces, so that the ROADM can be deployed as part of a disaggregated system where the various network elements may be provided by different vendors. In particular, our experimental validation shows that it is compatible with transponders from multiple vendors, and future work will likely show that the module is usable in fully disaggregated ROADMs as well.
In an open line system (OLS), the signals enter the network through the add stage of a ROADM. After traveling through several line degrees and long-distance fiber paths, possibly with in-line amplifiers, the signals eventually reach their destination and are sent to the client equipment via the ROADM drop stage. Historically, these add/drop modules typically performed at least some level of wavelength filtering, e.g., via fixed muxes/demuxes. A modern ROADM design typically employs a wavelength selective switch (WSS) for the add/drop in order to provide the desirable set of flexgrid, colorless, directionless, contentionless properties.
Notwithstanding certain limits on the number of allowed channels and their maximum power [4], the selectivity of coherent receivers can be leveraged in ROADM design. Eliminating the WSSs from the add/drop modules reduces costs and lessens the narrowing of the optical passband due to filtering [5]. Using power splitters/couplers in the add/drop ROADM modules is—to a certain extent—similar to filterless [6] or semi-filterless [7] networks, albeit on a different scale.
This paper targets a fully disaggregated [8], modular ROADM design where each module of a complete ROADM node is implemented via a standalone one-rack unit. A complete $ N $-degree/$ M $-add/drop ROADM therefore consists of $ N $ line degree modules, and $ M $ add/drop modules (Fig. 1).
Fig. 1. An example five-degree fully disaggregated ROADM node comprises seven independent modules: five line degree and two add/drop. In a fully disaggregated ROADM, each of the modules is independent. Our design simplifies the internals of the add/drop modules (orange inset, original WSS-based design shown), while the line degree modules (violet inset) remain unchanged.
In this paper, we continue our previously published work [9] on open ROADM designs based on the route-and-select architecture with twin WSSs per ROADM module. As a new contribution, we introduce a splitter/coupler add/drop module optimized for transmission of optical signals that utilize coherent detection.
The rest of this paper is structured as follows. In Section 2, we present a detailed overview of the internal architecture of this design with emphasis on the difference from the previously described WSS-based design. Results of performance verification under varying conditions are described in Section 3. Our results, current work, and future plans are summarized in Section 4.
2. ADD/DROP ROADM DESIGN
The ROADM designs created at CESNET, a Czech National Research and Educational Network (NREN), as a part of the Czech Light family [10], previously used the route-and-select architecture in both express and add/drop directions. The add/drop module of the ROADM was based on a 20-port WSS (for spectrum routing) and an optical channel monitor (OCM) for spectrum monitoring (see Fig. 2).
Table 1. Approximate Component Cost Structure for Different ROADM Modules
This design is flexible and compatible with alien wavelengths [11], but the required flexgrid WSS and OCM modules together constitute a major cost component of the entire device [12]. Thus, we designed an alternative open add/drop module that does not include these components, which we refer to as the coherent add/drop.
The approximate cost structure for CESNET’s open designs is provided in Table 1. Specific prices are slightly different because exact numbers are covered by non-disclosure agreements, but using close approximations, the bill-of-materials pricing of the proposed design saves more than 75% of the original flexgrid $ {\rm WSS} + {\rm OCM}$ add/drop design. As shown by the rough cost breakdown, the cost savings for the WSS and OCM are the key contributors.
To get a rough approximation of the overall cost savings in a full ROADM node, we also show the cost breakdown of a line degree module (see last column in Table 1). For a ROADM with five line degrees and two add/drop modules, the overall cost of the ROADM is reduced by roughly 25%.
When the client transponders are under full control of the OLS operator, a passive solution along with transponder selectivity is therefore a compelling alternative—if the performance remains acceptable, and if the lack of filtering and spectrum monitoring is compensated for by the control plane.
A. Optical Design
Utilizing power splitters and couplers in the add/drop module as opposed to WSSs is by no means novel [13–18]. Apart from spectrum filtering, an obvious advantage of a WSS-based add/drop design is a lower insertion loss (IL) of a WSS as compared to a passive, splitter-based architecture, and a potentially higher client port count. Typical splitter power losses are shown in Table 2.
Table 2. Typical Insertion Loss for a 1:N DWDM Splitter
A worst-case datasheet IL of a flexgrid WSS is quoted as 8 dB, and better solutions are yet to become commercially available [19]. The 11 dB IL of a 1:8 splitter is only 3 dB worse, and the losses are easily compensable by a node’s built-in amplification.
Figure 3 shows the internal optical schema of the proposed add/drop module. Compared to the original, WSS-based add/drop (Fig. 2), the twin WSS is replaced by a pair of 1:8 couplers, yielding a slightly worse IL. The OCM for channel monitoring in both add and drop directions is replaced by an integrated tap array with photodiodes (PDs) and a digital-to-analog converter (DAC) [20] in the add direction. The client port count is reduced from 20 WSS-based ports to just eight. The rest of the optical path remains unchanged, with the express-facing fan-out based on a pair of passive couplers, and a pair of erbium-doped fiber amplifiers (EDFAs) for loss compensation.
This add/drop design does not support channel monitoring in the drop direction, and channel equalization is not supported in either direction. In this regard, the design relies upon the capabilities of the line degree modules, i.e., the ingress line degree module within the same ROADM node, or the egress line degree module in the same ROADM node (refer to Fig. 1). Because the line degrees are still of the route-and-select architecture with flexgrid WSS in both directions (i.e., with filtering capabilities in both Express → Line OUT as well as Line IN → Express), the following features are still supported:
- • Only those media channels (MCs) that are to be dropped at a given add/drop module are routed towards that particular module. Under normal operation, the add/drop module will therefore receive one MC per connected transponder, i.e., up to eight MCs in total.
- • Monitor MCs that are dropped to a transponder through the add/drop module. Their power levels are measured by the built-in OCM at each corresponding ingress line degree’s express output. The filterless add/drop module monitors only aggregate power levels across all MCs.
- • Perform per-MC power equalization for the add direction. The added signals are broadcast to all egress line degrees, where a WSS between the egress line degree’s input and the line output can utilize all of its 15 dB effective dynamic range available for fine-tuning the MC signal power.
- • Prevent multi-path interference by adding only those MCs configured to egress over a given line degree’s Line OUT. In our design, this does not constitute any change compared to a ROADM with a WSS-based add/drop module.
The optical design of the line degree modules guarantees that signals at the outputs of the ingress line degrees are at a power level of at least $ - {12}\;{\rm dBm}$ per MC while accounting for the worst-case component tolerances [9]. For the coherent add/drop module, it is desirable to use the same EDFA module as in the line degrees [a twin-stage fixed-gain module, $ {+} {27}\;{\rm dB/ {+} 22}\;{\rm dB}$ automatic gain control (AGC) mode]. Each 1:8 power coupler attenuates by about 11 dB, and a pair of patch cords contributes about 0.5 dB each, leading to the minimal Rx power level at the transponder of $ - {8}\;{\rm dBm}$ (i.e., $ - {12}\;{\rm dBm} - {11}\;{\rm dB} + {27}\;{\rm dB} - {11}\;{\rm dB} - {1}\;{\rm dB}$).
A power level of $ - {8}\;{\rm dBm}$ at a receiver’s Rx would be slightly outside of the optimal Rx power range [21]. However, the original design of line degrees assumed a 4 dB loss in the dispersion compensation unit (DCU). Current revisions of line degrees remove the DCU, replacing it with a direct internal connection, thereby improving the power budget by 4 dB. During the initial prototyping, we also saw typical WSS performance and component tolerances contributing to a few more dB of margin. Actual operating conditions are typically better by at least 5 dB. Even assuming the worst-case tolerances and with component aging factored in, the design power level at the drop port is therefore at least $ - {4}\;{\rm dBm}$.
It is possible to use higher-port-count splitters (and therefore increase the number of client ports at one add/drop module). However, a 1:16 coupler exhibits 4 dB higher IL, and that penalty would shift the power budget into a non-optimal range at the transponder’s receiver port. Given this, the design is limited to an eight-port add/drop module.
With this limit on the number of channels that can be supported in each add/drop module, the design is best suited for metro applications, especially at small nodes, where the requirements for the number of channels to be added/dropped may not be very large. Otherwise, a very large number of add/drop modules would need to be deployed.
The first prototype device is shown in Fig. 4.
B. System Architecture
In terms of software and hardware, we kept the architecture of the system as a whole as close as possible to the system used for the line degree module as described in [9]. The coherent add/drop module once again exposes a NETCONF [22] interface for management and monitoring. YANG-formatted data [23] are managed by Sysrepo [24], and the NETCONF server is provided by the Netopeer2 [25] software. Custom code acts as an interface between Sysrepo’s callbacks and the optical hardware.
These callbacks operate over an internally abstracted property tree, loosely based on XPath [26] addresses of nodes within a YANG model. We wrote a new driver for the PD tap array, the only component not used in previous designs.
The whole system uses GNU/Linux with a modern user space software, including the systemd suite of services. The operating system images are built via Buildroot. The device’s internal flash memory is split into A/B firmware slots, and a hardware watchdog along with U-Boot bootloader integration ensures that the device automatically recovers upon failure scenarios, including an unresponsive user-space software or the operating system kernel.
During system burn-in, metadata such as module type (line degree, WSS add/drop, coherent add/drop) are stored in the bootloader environment along with Ethernet MAC addresses, etc. The bootloader passes these module parameters to the Linux kernel, and systemd starts an appropriate service. The code then initializes the expected devices, loads an appropriate YANG model, and starts servicing user requests over NETCONF. This infrastructure enables a unified software image for all ROADM modules regardless of their very different YANG interfaces.
Apart from the NETCONF interface, the system can also be managed via a locally running interactive console, the netconf-cli [27,28]. This program is accessible both over a serial-over-USB port at the front panel as well as remotely, over secure shell (SSH). The command-line interface (CLI) offers interactive <Tab>-key completion that is built during runtime based on the content of the YANG model (Fig. 5). The program accesses the common YANG-formatted data store via Sysrepo, and it therefore supports the same set of features as any NETCONF client. The CLI NETCONF console was released as an open source project under the terms of the Apache-2 license.
Listing 1. Tree View of the Add/Drop YANG Model Subset
C. YANG Model Monitoring and Telemetry
While many of the ROADM add/drop features are the same, there are some important differences when the WSS is replaced by a coupler/splitter and the OCM by a PD array. The YANG model is shown in Listing 1. The device does not offer any hardware control on whether signals are passed, blocked, or attenuated. (Technically, the EDFA could be switched off and on again, but there is no real use case for that during operation. The code therefore configures the EDFA to a correct operating mode, AGC, and then monitors its operating parameters.) The majority of the YANG model therefore focuses on retrieving information about optical power, and on alarm threshold configuration. The only other writable item is a per-port label with user-defined semantics.
As an improvement to the operator-facing user interface, we added eight LEDs to indicate the state of each add port (see Fig. 4) directly on the equipment front panel. After configuring power thresholds via NETCONF, a local software loop continuously monitors signal levels and indicates them via dual-color red–green LEDs. The same data are also available over the northbound NETCONF interface. The model supports power monitoring at each ingress client port (add) with configurable alarm thresholds and hysteresis. There is an aggregate power monitoring on all multi-wavelength physical connectors—in this case, Express IN, Express OUT, and all client drop ports, and the related YANG structure was reused from the original ROADM with WSS-based add/drops.
An important difference compared to the WSS-based add/drop module is the use of a PD tap array for client port monitoring instead of a flexgrid OCM. While the OCM provides detailed spectrum information, a typical measurement cycle takes 300 ms, leading to about 3 Hz of update frequency. The PD tap array, instead, offers measurement latencies in milliseconds. We slowed down the software readout process to run at 20 Hz update frequency, still promising excellent performance for upcoming streaming telemetry work. The existing literature typically argues for 1 s sampling [29–31]. However, fast monitoring, possibly even without streaming the telemetry to a centralized controller, could enable machine-learning-based analysis of MC performance [32].
In our previous work, we used NETCONF notifications for per-MC statistics and used them as an input for a simple software-defined network (SDN) demonstration of optical path protection [33]. Unfortunately, the implementation of notification handling in the currently released version of Sysrepo suffers from a serious performance degradation during high-volume notification traffic [34], and also during transfers of large list instances [35]. A new version of Sysrepo where these performance-related shortcomings were addressed has not been released yet, and NETCONF does not appear to be a suitable transport channel for frequent, periodic transfers of bulk data. Other researchers also resorted to throttling of NETCONF notifications to send them, e.g., every 15 min [32].
The industry consensus appears to have formed on gRPC [36] and gNMI [37] for actual data transport [29]. The code will be able to pass the performance numbers obtained from the hardware to a future gNMI backend. A new Sysrepo architecture with operational data caching looks especially promising. We have disabled NETCONF notifications in the meanwhile as a stopgap measure.
3. PERFORMANCE MEASUREMENT
The available literature contains results of many experiments related to interactions between multiple channels with possibly different modulations [38–40]. Even when we disregard different modulation schemas and restrict the considered operating envelope to only those utilizing the same dual-polarization quadrature phase-shift keying (DP-QPSK) modulation, the number of present channels still affects the performance of the original signal [41]. This section compares their contribution to the pre-forward-error-correction (FEC) bit error rate (BER) under various power levels of the “parasitic” signals.
Fig. 6. Schema of the experimental setup for performance validation. L1, L2, and L3 are line degree modules.
During the performance evaluation, we used one to four channels from two Coriant (now Infinera) Groove G30 flexible transponders set at 100 Gbps DP-QPSK modulation, and measured their pre-FEC BER rates against the receiver-reported Rx per-channel power [42].
We did not want to conduct a simple back-to-back testing, so the actual fiber path included two 100 km spans of G.652.D fiber and two additional line degree ROADM modules in an attempt at simulating less sterile signal conditions [43]. The setup is shown in Fig. 6. The signals were muxed in via a “Coherent A/D,” crossed the egress stage of the “L1” line degree module, sent to a 100 km fiber span, received at “L2,” looped back via “L2” into another 100 km fiber span, received by the ingress “L1,” and demuxed via the “Coherent A/D” before terminating at the other transponder. We varied the channel power available at the Rx port via the closest present WSS, i.e., the line degree module immediately preceding the measured add/drop prototype—L1 in Fig. 6.
As an additional backfill for loading the coherent add/drop module with additional channels, we added a copy of three signals from the CESNET production network. These signals were tapped, passively muxed, and sent to the experimental setup with no additional processing. We amplified them using another ROADM line degree module (L3) and brought them to the add/drop’s Express IN ports to simulate extra dropped traffic coming from another direction. If used, the absolute power of these three backfill channels at the drop ports was fixed at 0 dBm per channel during all of the measurements to simulate worst-case conditions with non-optimized channel equalization. In a real scenario, these channels would likely be attenuated by a preceding line degree module (L3) so that all MCs dropped via the single coherent add/drop are power-equalized.
As explained in Section 2.A, during regular operation, there will never be more than eight channels at each coherent add/drop’s Express IN ports. The line degree WSSs on the ingress side of the ROADM will direct only the desired drop channels to the drop modules. This is why we did not simulate with a fully loaded C-band (96 channels). Our investigation focused only on performance penalties related to the presence of multiple signals at the receiver, not on nonlinear optical impairments in the fiber.
A. Results and Discussion
Figure 7 shows our results. In all of the traces, each respective Coriant channel is consistently represented by the same graphical symbol (+, ×, ★, ▪). The color indicates the measurement configuration of the add/drop—one channel alone to establish a baseline (blue), two adjacent channels (orange), four channels (green), or seven channels (red). (The seven-channel configuration is described below.)
The blue traces exhibit an interesting variation between each other, indicating that even in a single-channel configuration with no other channels present, there were significant differences in the observed signal quality depending on which of the four channels (A, B, C, or D) was being dropped. This observation underlines the importance of testing under different operating conditions.
The general trend across all tested transponder configurations is clearly visible—the more signals present at the transponder Rx port, the bigger the impairment impact on the signal quality. As mentioned above, due to the WSS present in all of the ROADM line degree modules, the design guarantees that under normal operation, there will never be more than eight channels present on any coherent add/drop’s Express IN ports, preventing a possible Rx overload and further signal deterioration.
B. Transmit Channel Filtering and Tx Band Shaping
An important consideration when using WSS-less add/drop is transmitter channel filtering [44]. In our next measurement, we show three older-generation Cisco DP-QPSK 100 Gbps channels with no Tx band shaping. Their modulation therefore partially impairs the spectra of Coriant’s signals [45]. The spectral overlap of the adjacent unfiltered channels as captured on a high-resolution optical spectrum analyzer (OSA) is shown in Fig. 8. The blue OSA trace shows the aggregate spectra of four signals from the Coriant transponders; these transponders employ Tx band shaping and therefore do not exhibit any significant crosstalk among individual channels even with no guard bands and no WSS-based filtering. The green trace, recording spectra of three unfiltered Cisco channels, however, shows significant ($ - {7}\;{\rm dB}$ relative to the signal peak) crosstalk, which leaks to the adjacent channel.
Fig. 8. OSA trace of unfiltered transmitters shows the risk of spectrum incursions to the adjacent channels.
We deliberately configured the WSS in line degree module L3 to pass the whole spectrum to the coherent add/drop’s Express IN. Figure 9 shows the resulting pre-FEC BER deterioration. The only varying factor between the traces are unshaped Cisco channels (gray) versus WSS-filtered Cisco signals (red, same as in Fig. 7). The Rx power of each unfiltered Cisco MC was set to 0 dBm per MC to simulate a worst-case scenario.
4. CONCLUSION
We presented an open design of the add/drop ROADM module that extends our previously published work on OLSs. The proposed design provides significant cost savings on the add/drop modules and reduces the impact of filter narrowing due to WSS cascades due to the elimination of the WSSs from the add/drop modules. It is designed for carrying signals from coherent transponders that are under control of the network owner (i.e., not alien wavelengths). As the module offers only eight client ports, it is suitable for scenarios where the total number of dropped channels is limited. The software side of the device exports all measured optical properties of all client signals via a NETCONF interface to an SDN controller and a local CLI.
Measurements show that the performance impact of a WSS-free drop stage depends on the number of concurrently dropped channels and their equalization. The total impairment, however, is comparable with regular variations between individual channels and pieces of equipment. Transmitter signal shaping or use of guard bands remains an important consideration for preventing channel excursion from adjacent, possibly misbehaving, client signals. In many scenarios, signals are filtered by the immediately preceding line degree ROADM module.
A. Experimental Validation
Our concept of completely disaggregated ROADMs, where each line degree and each add/drop shared risk group is implemented by a standalone unit with its own YANG data store available over a dedicated NETCONF endpoint, poses certain challenges. The filterless ROADM design proposed in this paper does not have any hardware capable of power equalization or spectrum filtering, for example.
To prove that our approach to full disaggregation is viable, we engaged with the Open Networking Foundation (ONF)’s Open Disaggregated Transport Network (ODTN) [46] and the Telecom Infra Project (TIP)’s Open Optical Packet Transport (OOPT) groups. We contributed ROADM device drivers for Czech Light ROADM line degrees, the WSS-based add/drop modules, and also the coherent add/drop module described in this paper to ONF’s Open Network Operating System (ONOS). It was possible to adapt ODTN’s approach for provisioning individual MCs as FlowEntry and FlowRule instances to our ROADM design with no changes to the ONOS core.
To showcase interoperability, we simulated the setup with a WSS-less add/drop in GNPy [47], an open source dense wavelength division multiplexing (DWDM) simulation engine. An interactive demo with end-to-end path provisioning over fully disaggregated OLS was presented at the TIP Summit 2019 in Amsterdam. An improved version of the demo was accepted for presentation at OFC 2020 [48], and will be described in a follow-up article.
B. Future Work
Current capabilities of the coherent add/drop ROADM module are limited by the wider software ecosystem support of streaming telemetry. The upcoming release of Sysrepo (in development as of December 2019) shows promising performance improvements. Bridging Sysrepo’s YANG data store and operational data callbacks with a gNMI/gRPC ecosystem will allow low-latency bulk data streaming with low overhead and sub-50 ms update latencies.
Similarly, optical layer support in ONOS can be extended. This device’s fast performance readout can provide a valuable stream of data towards the SDN controller.
Funding
Ministerstvo Školství, Mládeže a Tělovýchovy (CZ.02.1.01/0.0/0.0/16_013/0001797); Telecom Infra Project (OOPT-PSE).
Acknowledgment
The authors would like to thank Jakub Mer, Michal Hažlinský, Esther Le Rouzic, and Andrea Campanella for their valuable contribution towards this paper.
REFERENCES
1. K. Kikuchi, “Coherent transmission systems,” in 34th European Conference on Optical Communication (2008), pp. 1–39.
2. L. Kazovsky, “Multichannel coherent optical communications systems,” J. Lightwave Technol.5, 1095–1102 (1987). [CrossRef]
3. M. O’Sullivan, “Expanding network applications with coherent detection,” in Optical Fiber Communication Conference and the National Fiber Optic Engineers Conference (OFC/NFOEC) (2008), pp. 1–17.
4. C. Xia, W. Schairer, A. Striegler, L. Rapp, M. Kuschnerov, J. F. Pina, and D. van den Borne, “Impact of channel count and PMD on polarization-multiplexed QPSK transmission,” J. Lightwave Technol.29, 3223–3229 (2011). [CrossRef]
5. M. Filer and S. Tibuleac, “N-degree ROADM architecture comparison: broadcast-and-select versus route-and-select in 120 Gb/s DP-QPSK transmission systems,” in Optical Fiber Communication Conference (2014), pp. 1–3.
6. G. Mantelet, C. Tremblay, D. V. Plant, P. Littlewood, and M. P. Bélanger, “PCE-based centralized control plane for filterless networks,” IEEE Commun. Mag.51(5), 128–135 (2013). [CrossRef]
7. S. Khanmohamadi, J. Chen, F. Abtahi, L. Wosinska, A. Cassidy, E. Archambault, C. Tremblay, S. Asselin, P. Littlewood, and M. Bélanger, “Semi-filterless optical network: a cost-efficient passive wide area network solution with effective resource utilization,” in Asia Communications and Photonics Conference and Exhibition (ACP) (2011), pp. 1–3.
8. E. Riccardi, P. Gunning, O. G. de Dios, M. Quagliotti, V. López, and A. Lord, “An operator view on the introduction of white boxes into optical networks,” J. Lightwave Technol.36, 3062–3072 (2018). [CrossRef]
9. J. Kundrát, O. Havliš, J. Jedlinský, and J. Vojtěch, “Opening up ROADMs: let us build a disaggregated open optical line system,”J. Lightwave Technol.37, 4041–4051 (2019). [CrossRef]
10. “Czech Light open line system,” 2019, https://czechlight.cesnet.cz/en/open-line-system/sdn-roadm.
11. H. Wessing, P. Skoda, M. N. Petersen, A. Pilimon, P. Rydlichowski, G. Roberts, R. Smets, J. Radil, J. Vojtech, R. Lund, Z. Zhou, C. Tziouvaras, and K. Bozorgebrahimi, “Alien wavelengths in national research and education network infrastructures based on open line systems: challenges and opportunities,” J. Opt. Commun. Netw.11, 118–129 (2019). [CrossRef]
12. P. Papanikolaou, P. Soumplis, K. Manousakis, G. Papadimitriou, G. Ellinas, K. Christodoulopoulos, and E. Varvarigos, “Minimizing energy and cost in fixed-grid and flex-grid networks,” J. Opt. Commun. Netw.7, 337–351 (2015). [CrossRef]
13. S. Gringeri, B. Basch, V. Shukla, R. Egorov, and T. J. Xia, “Flexible architectures for optical transport nodes and networks,” IEEE Commun. Mag.48(7), 40–50 (2010). [CrossRef]
14. M. Nooruzzaman and E. Halima, “Low-cost hybrid ROADM architectures for scalable C/DWDM metro networks,” IEEE Commun. Mag.54(8), 153–161 (2016). [CrossRef]
15. G. Zervas, E. Hugues-Salas, T. Polity, S. Frigerio, and K.-I. Sato, Node Architectures for Elastic and Flexible Optical Networks (Springer, 2016).
16. W. Jin, C. F. Zhang, X. L. Zhang, X. Duan, Y. X. Dong, R. P. Giddings, K. Qiu, and J. M. Tang, “OSNR penalty-free add/drop performance of DSP-enabled ROADMs in coherent systems,” J. Opt. Commun. Netw.9, 730–738 (2017). [CrossRef]
17. W. J. Jiang, A. Lebedev, Y. M. Lin, I. Vorobeichik, I. Gopp, A. Plotskiy, and W. I. Way, “Degree-expandable colorless, directionless, and contentionless ROADM without drop-side EDFAs,” in Optical Fiber Communication Conference and Exhibition (OFC) (2015), pp. 1–3.
18. D. G. Sequeira, L. G. Cancela, and J. L. Rebola, “Impact of physical layer impairments on multi-degree CDC ROADM-based optical networks,” in International Conference on Optical Network Design and Modeling (ONDM) (2018), pp. 94–99.
19. N. K. Fontaine, R. Ryf, D. T. Neilson, and H. Chen, “Low loss wavelength selective switch with 15-THz bandwidth,” in European Conference on Optical Communication (ECOC) (2018), pp. 1–3.
20. Oplink, “Tap PD array with I2C, single-mode,” 2019, http://www.oplink.com/passive/detail/110.html.
21. G. Roberts, “TIP Open Optical Packet Transport—a game-changer for R&E networking,” 2017, https://www.cesnet.cz/wp-content/uploads/2017/05/3_Guy_Roberts_TIP-Open-Optical-Packet-Transport-A-game-changer-for-RE-networking.pdf.
22. R. Enns, M. Björklund, A. Bierman, and J. Schönwälder, “Network Configuration Protocol (NETCONF),” IETF RFC 6241 (2011).
23. M. Björklund, “The YANG 1.1 data modeling language,” IETF RFC 7950 (2016).
24. “Sysrepo—YANG-based configuration and operational state data store for Unix/Linux applications,” 2019, http://www.sysrepo.org/.
25. “Netopeer2—The NETCONF Toolset,” Git repository, 2019, https://github.com/CESNET/Netopeer2.
26. World Wide Web Consortium and others, “XML Path Language (XPath), 1.0,” W3C Recommendation, 1999.
27. V. Kubernát, “Nástroj pro konfiguraci a monitorování,” B.Sc. thesis (Faculty of Information Technology, Czech Technical University in Prague, 2019), https://dspace.cvut.cz/handle/10467/83195.
28. “Console interface to NETCONF servers,” 2019, https://gerrit.cesnet.cz/plugins/gitiles/CzechLight/netconf-cli/.
29. F. Paolucci, A. Sgambelluri, F. Cugini, and P. Castoldi, “Network telemetry streaming services in SDN-based disaggregated optical networks,” J. Lightwave Technol.36, 3142–3149 (2018). [CrossRef]
30. L. Velasco, A. C. Piat, O. Gonzlez, A. Lord, A. Napoli, P. Layec, D. Rafique, A. D’Errico, D. King, M. Ruiz, F. Cugini, and R. Casellas, “Monitoring and data analytics for optical networking: benefits, architectures, and use cases,” IEEE Netw.33, 100–108 (2019). [CrossRef]
31. J. Kawasaki, S. Sumita, G. Mouri, M. Miyazawa, and T. Tsuritani, “Versatile optical path calculation and monitoring using common YANG model,” in IFIP/IEEE Symposium on Integrated Network and Service Management (IM) (2019), pp. 563–568.
32. A. Sgambelluri, J. Izquierdo-Zaragoza, A. Giorgetti, L. Gifre, L. Velasco, F. Paolucci, N. Sambo, F. Fresi, P. Castoldi, A. C. Piat, R. Morro, E. Riccardi, A. D’Errico, and F. Cugini, “Fully disaggregated ROADM white box with NETCONF/YANG control, telemetry, and machine learning-based monitoring,” in Optical Fiber Communications Conference and Exposition (OFC) (2018), pp. 1–3.
33. J. Kundrát, J. Vojtěch, P. Škoda, R. Vohnout, J. Radil, and O. Havliš, “YANG/NETCONF ROADM: evolving open DWDM toward SDN applications,” J. Lightwave Technol.36, 3105–3114 (2018). [CrossRef]
34. “Sysrepo issue 735: poor performance of notifications,” 2019, https://github.com/sysrepo/sysrepo/issues/735.
35. “Sysrepo issue 1635: returning data from RPC: very slow operation on lists with 5k+ items,” 2019, https://github.com/sysrepo/sysrepo/issues/1635.
36. “gRPC, a high performance, open-source universal RPC framework,” 2015, https://grpc.io/.
37. R. Shakir, “gNMI overview,” 2018, https://datatracker.ietf.org/meeting/101/materials/slides-101-netconf-grpc-network-management-interface-gnmi-00.
38. M. Birk, P. Gerard, R. Curto, L. E. Nelson, X. Zhou, P. Magill, T. J. Schmidt, C. Malouin, B. Zhang, E. Ibragimov, S. Khatana, M. Glavanovic, R. Lofland, R. Marcoccia, R. Saunders, G. Nicholl, M. Nowell, and F. Forghieri, “Real-time single-carrier coherent 100 Gb/s PM-QPSK field trial,” J. Lightwave Technol.29, 417–425 (2011). [CrossRef]
39. R. Borkowski, F. Karinou, M. Angelou, V. Arlunno, D. Zibar, D. Klonidis, N. G. Gonzalez, A. Caballero, I. Tomkos, and I. T. Monroy, “Experimental study on OSNR requirements for spectrum-flexible optical networks [Invited],” J. Opt. Commun. Netw.4, B85–B93 (2012). [CrossRef]
40. F. Cugini, C. Porzi, N. Sambo, A. Bogoni, and P. Castoldi, “Receiver architecture with filter for power-efficient drop waste networks,” in Optical Fiber Communication Conference and Exhibition (OFC) (2016), pp. 1–3.
41. B. Zhang, “Advanced receiver design for next generation coherent optical system,” in 13th International Conference on Optical Communications and Networks (ICOCN) (2014), pp. 1–4.
42. T. Uchikata, Y. Suzuki, and A. Noda, “A high accuracy channel power extraction method in optical filter-less coherent detection for flexible ROADM,” in 17th Opto-Electronics and Communications Conference (2012), pp. 761–762.
43. M. Filer, M. Cantono, A. Ferrari, G. Grammel, G. Galimberti, and V. Curri, “Multi-vendor experimental validation of an open source QoT estimator for optical networks,” J. Lightwave Technol.36, 3073–3082 (2018). [CrossRef]
44. D. Gariépy, S. Searcy, G. He, S. Tibuleac, M. Leclerc, and P. Gosselin-Badaroudine, “Novel OSNR measurement techniques based on optical spectrum analysis and their application to coherent-detection systems,” J. Lightwave Technol.37, 562–570 (2019). [CrossRef]
45. P. Castoldi, T. Foggi, F. Paolucci, and F. Cugini, “Next steps in elasticity: enabling signal overlap in optical networks,” in 18th International Conference on Transparent Optical Networks (ICTON) (2016), pp. 1–4.
46. A. Campanella, H. Okui, A. Mayoral, D. Kashiwa, O. G. de Dios, D. Verchere, Q. Pham Van, A. Giorgetti, R. Casellas, R. Morro, and L. Ong, “ODTN: Open Disaggregated Transport Network. Discovery and control of a disaggregated optical network through open source software and open APIs,” in Optical Fiber Communication Conference and Exhibition (OFC) (2019), pp. 1–3.
47. A. Ferrari, M. Filer, K. Balasubramanian, Y. Yin, E. Le Rouzic, J. Kundrát, G. Grammel, G. Galimberti, and V. Curri, “Experimental validation of an open source quality of transmission estimator for open optical networks,” paper accepted for OFC 2020.
48. J. Kundrát, A. Campanella, E. Le Rouzic, A. Ferrari, O. Havliš, M. Hažlinský, G. Grammel, G. Galimberti, and V. Curri, “Physical-layer awareness: GNPy and ONOS for end-to-end circuits in disaggregated networks,” demo accepted for OFC 2020.
