Open Access
Add to BibSonomyAbstract
The surge in Internet usage and connectivity worldwide coupled with the emergence of 5G mobile communications has amplified the need for a scalable and sustainable telco networking model. This paper outlines our Open Broadband trial in Brazil and the journey to open software and whitebox hardware as a practical approach to scaling fiber access networks enabling integration into a software-defined network controlled environment while transitioning away from vertically integrated models featuring black box passive optical network (PON) solutions. A key part of this journey is multi-PON capable whitebox optical line terminals (OLTs) with combo-PON (C-PON) optics that play a key role in bridging the transition from G-PON to XGS-PON.
© 2021 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. INTRODUCTION
As bandwidth demands driven by digitization and ubiquitous connectivity continue to increase, the telecom industry is challenged to build and scale their networks in a sustainable manner. Traditional network architectures are currently facing a scalability and flexibility challenge in the era of hyperconnectivity where millions of devices are continuously added to the network. A recent example is the unexpected crisis caused by the COVID-19 pandemic, which caused a sudden surge in broadband usage globally, spiking the number of connected devices online during the workday. On March 11, 2020, a nationwide lockdown was implemented in the United States, causing a drastic increase in broadband usage overnight that has sustained throughout the pandemic. The nationwide lockdown, which mandated virtual learning for students and telecommuting as much as possible, resulted in 42.8 million active devices online (from homes) during the workday. That is a 90% increase from pre-pandemic times when 22.6 million devices were typically active online (from homes) during the workday [1]. See Table 1 for a list of acronyms used in this paper.
Table 1. List of Acronyms
Spain, for example, experienced a nearly 40% sudden increase in Internet protocol (IP) traffic within a few days at the onset of the COVID-19 crisis due to the massive and simultaneous use of broadband networks for remote work, virtual learning, online entertainment, and instant messaging during business hours.
Similarly, throughout the pandemic cable Internet networks have also experienced a steady and sustained increase in downstream peak traffic with 25.5% cumulative growth since March 1, 2020 [2]. Consistent with that trend, upstream peak growth also continues to display dynamic surges up a total of 49.4% since that date [2].
By 2023, it is forecasted that global Internet adoption and connectivity will be so pervasive that nearly two-thirds of the global population will have Internet access, and the number of devices connected to IP networks will be more than three times the global population [3]. The share of machine-to-machine (M2M) connections is expected to grow to 50% by 2023 with 14.7 billion M2M connections.
Simultaneously the fifth generation of mobile communications (5G) is burgeoning, welcoming new innovations and applications, such as edge computing. The market is already seeing pent-up demand for 5G innovations as network users, primarily businesses, demand personalized and low-latency services that can be provisioned in real-time. To meet the challenges of this current market and the foreseeable future, network operators need flexible, high-performance, real-time responsive network architectures that can be deployed quickly in a cost-effective, sustainable manner.
In a recently released paper [4], Telefónica reported the results of a software-defined passive optical network (SD-PON) trial where 10 gigabit capable symmetric PON (XGS-PON) technology was used in a stand-alone configuration compatible with legacy network architectures to allow fast, cost-effective deployments. In this article, we report the results of a new SD-PON trial using a whitebox OLT with flexible PON technology and describe our proposal for what we believe, to the best of our knowledge, is a new and open combo-PON (C-PON) optical transceiver that allows a smooth evolution from G-PON to XGS-PON.
This paper has six sections. First, we introduce the framework and the principles of our joint vision on open and disaggregated fixed access networks in Section 2. In Section 3, we describe our proposal for an open and disaggregated software architecture with flexible centralized/distributed deployment of the control plane for PON access networks and provide the technical insights on how the SD-PON software is optimized. In Section 4, we report the first results of our latest trial using a whitebox optical line terminal (OLT) with flexible PON technology. The evolution of fiber-to-the-home (FTTH) optics toward combo-PON technology and our proposal for a new and open combo-PON optics transceiver is presented in Section 5. Finally, the conclusions of this paper and our plans for future work are described in Section 6.
2. JOURNEY TOWARD OPEN BROADBAND IN FIXED NETWORKS
Optical line terminals are the access nodes for FTTH networks and are indispensable components of PONs. In the case of FTTH PON, the OLTs connect with the customer premises equipment (CPE), called the optical networking unit/termination (ONT/U) using a branched optical distribution network with passive optical splitters. In some scenarios, CPEs may comprise an XGS-PON ONT and a Wi-Fi gateway integrated in a single device, also known as XGS-PON home gateway units (XHGUs).
In legacy network architectures, OLT hardware is so tightly coupled with the software that it has slowed down operators’ network evolution to the pace of a single PON vendor’s roadmap and the development of customized services. In this type of closed and monolithic architecture, hardware changes or upgrades can be operationally disruptive, inefficient, and costly to implement. Bringing in new innovative features or functions is now dependent on the PON vendor. To resolve these issues for network operators and overcome these limitations, several industry groups, such as the Open Networking Foundation (ONF), the Open Compute Project (OCP), and the Telecom Infra Project (TIP), as well as standardization organizations like the Broadband Forum (BBF), developed architectures in recent years to transform legacy networks, migrating them toward software-defined networks (SDNs) and network function virtualization (NFV) using cloud technologies.
Implementing key learnings from these industry groups, Telefónica’s Open Broadband (OBB) initiative is the evolution of its PON access network based on hardware and software disaggregation, SDN, and cloud computing technologies. The guiding principles of Telefónica’s Open Broadband initiative are as follows:
- • Disaggregate the software from the access node hardware. With platform-agnostic software, this approach allows the software for switching, routing, and the control plane of a whitebox OLT to be designed independent of the specific hardware platform.
- • Commoditize hardware components and customized design. The physical network functions (PNFs) for access and aggregation equipment are built based on commercially available merchant silicon, also known as whiteboxes, with design flexibility and opportunity for customization.
- • A general-purpose processor (GPP) architecture enables the software to be deployed in virtualized environments, allowing convergence with 5G mobile networks, edge computing applications, and future virtual network functions (VNFs).
Figure 1 depicts the Open Broadband architecture with access nodes formed by whitebox OLTs and top of rack (ToR) switch routers for aggregation arranged in a cloud-native central office (Cloud CO) [5] domain with multi-access edge computing (MEC) services, VNFs, and third-party applications. As part of the Open Broadband initiative, Telefónica demonstrated at the Mobile World Congress 2019 a converged fixed and mobile network sharing the same virtualization infrastructure for the network control plane and MEC applications for FTTH customer service auto-provisioning and real-time facial image processing [6].
Fig. 1. Open Broadband multi-PON virtual OLT scenario for fixed network evolution (EMS, element management system; OSS, operation and support system; MANO, management and orchestration).
3. SOFTWARE VIRTUALIZATION AS AN ENABLER OF HARDWARE AND SOFTWARE DISAGGREGATION
Virtualization technologies enable any software to run ubiquitously and scale across physical boundaries (e.g., servers). These technologies are well suited to optimize network functions with fluctuating resource usage and frequent modifications due to evolving standards and network security requirements.
The disaggregated PON control and management software in Telefónica’s Open Broadband initiative is built using a generic layered architecture, which uses adaptation layers on northbound and southbound interfaces (NBIs and SBIs). That design enables flexibility to adapt to a multitude of protocols (G-PON, XGS-PON/NG-PON2, 25 Gb/s PON, 50 Gb/s PON, 100 Gb/s PON) and management mechanisms, such as management and orchestration (MANO), traditional operational support systems/business support systems (OSSs/BSSs), and PON control and management, which are responsible for the business logic of the software stack.
The virtual ONU management and control interface (vOMCI) from the Broadband Forum in the Open Broadband–Broadband Access Abstraction (OB–BAA) project [7] and the virtual OLT hardware abstraction (VOLTHA) from the ONF [8] are examples of a virtualized software stack for whitebox OLTs.
The SD-PON system as depicted in Fig. 2 and presented in this paper is built on the ONF’s VOLTHA project [8]. The software is organized into microservices, each of which has a very specific function and is an independent component of the software stack. Using a cloud-native microservice architecture, the design brings scalability in service and software upgrades and resiliency against faults in the virtualization platform and software. This architecture allows upgrades to the individual components without affecting the larger system. The important differentiator is having key interfaces that are open; open as in “defined by a community of players across the industry, including operators, vendors, systems integrators, and major component suppliers (merchant silicon).” One such interface is the whitebox OLT interface (Fig. 2) defined by the ONF’s SEBA/VOLTHA community. The whitebox OLT interface defines a clean and clear demarcation of the PON control and management software (SD-PON) from the underlying PON TC/MAC/physical layers implemented as merchant silicon in the hardware (whitebox OLT). One significant benefit is that it allows for a larger ecosystem (versus a limited set of proprietary OLT vendors) of software and hardware suppliers, which in turn allows for greater innovation to be tapped by operators. The same applies to interfaces higher up in the stack (Fig. 2) between the SD-PON and element management system (EMS) as well as the EMS and OSS/BSS/MANO platforms.
The ONF’s VOLTHA provides common vendor-agnostic PON control and management of whitebox OLTs. It abstracts the PON functionality and presents the underlying OLT and the ONTs as a simple programmable SDN switch.
With VOLTHA at the heart of the system, there are three types of microservices that comprise the complete SD-PON system: control, infrastructure, and management.
As depicted in Fig. 3, VOLTHA and an SDN controller are categorized as “control” microservices. The SDN controller performs the following functions: a dynamic host configuration protocol (DHCP) relay agent for IPv4 and IPv6, Internet Group Multicast Protocol (IGMP) proxy, flow configuration for subscriber services, and multicast flows across the whitebox OLT and ONUs.
The infrastructure microservices provide common platform functions like database (Redis), messaging and logging (Kafka), and container orchestration (Kubernetes).
The final set of microservices are the management microservices, which allow a northbound EMS/OSS to manage the SD-PON system.
There are five key management microservices:
- a) The EMS gateway acts as the interface toward a northbound EMS. All requests from the EMS are handled from here and routed to the appropriate microservice from completion. The EMS gateway provides an API and reports alarms, events, and other notifications on the Kafka message bus.
- b) The subscriber manager (SubMgr) handles all functions related to the subscribers and their services. Key functions include authorizing and activating subscribers, and subscriber profile management (including VLAN, QoS, and bandwidth).
- c) The device manager (DevMgr) manages the OLTs, ONTs, the PON, and NNI ports, providing the ability to add, delete, and update OLTs, ONTs, and PON/NNI ports. The device manager also monitors their operational state, reporting alarms from OLTs and handling blacklisted (unauthorized) ONTs.
- d) The network services manager (NwMgr) handles the L2 functions on the NNI interface. Key network services like E-Line or E-LAN, and ERPS are supported.
- e) The telemetry gateway provides a centralized data store (alarms, events, counters), which serves as the single source for external reporting.
Kubernetes is the container orchestrator and life cycle manager of the SD-PON software.
With the SD-PON software essentially running in an x86 environment, there are two types of deployment models—centralized and distributed.
The centralized deployment model [Fig. 4(A)] allows the SD-PON software to be executed on an x86 server cluster in a telco data center environment or in a public cloud (e.g., the Google Kubernetes engine on the Google Cloud Platform). This model is suitable for high-density locations (big cities and states) where the operator needs to support a greater number of subscribers.
As the centralized SD-PON deployment model runs on commercial off-the-shelf (COTS) hardware, server costs can be amortized across large numbers of subscribers; however, availability and reuse of existing data center compute resources allow for smaller scale centralized SD-PON deployments without incremental compute server costs. From a management perspective, a centrally located EMS instance (potentially collocated with the SD-PON instance) can manage all whitebox OLTs.
Alternatively, when the deployment density is low (such as in rural locations), locally available cloud computing resources might not be readily available or economically viable for single instance OLT deployments. For this use case, the distributed deployment model allows the SD-PON software to be executed on the whitebox OLT itself, which typically has just enough onboard computing and memory resources, like an Intel 8 Core central processing unit (CPU) and 64 GB random access memory (RAM) on a Radisys whitebox OLT.
The SD-PON software can execute on limited compute resources on the OLT itself as the compute resources are engineered to support the maximum number of subscribers per OLT. From a management perspective, a centrally located EMS instance shall manage all whitebox OLTs (with an embedded SD-PON).
The SD-PON software, designed to run in a central/cloud infrastructure, had to be optimized to fit into the limited footprint in terms of the CPU and RAM available on the OLT for the distributed deployment model. In the first attempt, described in an earlier paper [4], the Edgecore XGS-PON OLT’s RAM was upgraded to 32 GB and the solid-state drive (SSD) was upgraded to 256 GB. Based on this learning, in this trial the Radisys XGS-PON OLT has been designed with 64 GB RAM and 256 GB of SSD. The following software optimizations helped fit the SD-PON into the Radisys whitebox OLT:
- • The SDN controller that is part of the SD-PON stack is optimized for a smaller RAM footprint and lesser CPU consumption. While greatly improving the CPU consumption, it also added more responsiveness from the controller for applications such as DHCP relay and IGMP proxy, which must be processed in real-time. This enables completion of about 2000 DHCP negotiations within the span of 1 to 2 s.
- • The database, originally a purely disk-based system with high consistency, has been modified to be a combination of disk and RAM to improve the CPU consumption while not losing the benefits of disk-based consistency. This move reduced the CPU utilization of the database to a meager 10% of the original consumption.
- • A light common framework is the base for all microservices that eventually lead to a smaller footprints and better performance.
While the software was optimized across the board to fit into an OLT, the Kubernetes was retained to provide the benefits of cloud native software that allows automated rolling software updates that greatly reduce the cost and operational complexity experienced by the operators. The benefits of Kubernetes far outweigh the already small resource footprint it occupies.
A key benefit of the distributed model is its quick time-to-market (TTM). Since it operates very similarly to the traditional proprietary OLTs that most operators’ operational personnel are trained to deploy and maintain, it gives the operators full flexibility to mix and match the best PON control and management software with the best whitebox OLT hardware.
We believe the transition toward disaggregated access networks will be a journey starting with the distributed deployment model to gain experience and confidence, then slowly evolving to an all-centralized or a hybrid model with coexistence of both distributed and centralized deployment models. In the scenario where an operator chooses to migrate to an all-centralized model, the migration from the distributed to the centralized deployment model frees up computing resources on the OLTs, allowing for support of edge computing applications.
As shown in Fig. 5, we envision an eventual evolution of the OLT as an important part of the converged edge that may support other applications like a content delivery network (CDN), or a 5GC user plane function (UPF; network function) or any value-added applications (cloud gaming or augmented/virtual reality).
The evolution to a converged edge is expected to be a gradual process starting with limited initial deployments of the distributed control plane with trials in low-density deployments first, as shown in Fig. 4(A), and the centralized control plane with trials in high density deployments later, as illustrated in Fig. 4(B). The goal is to enable maximum flexibility for the network operator starting with PON/OLT disaggregation and virtualization that will be further enhanced to usher in a converged edge paradigm, as shown in Fig. 5.
The longer-term evolution will have two distinct themes: core network disaggregation and convergence and introduction of non-access network functions. Core network disaggregation and convergence involves broadband network gateway (BNG) disaggregation (wireline only operators) and converged 5G core (operators with both wireline and wireless networks). Non-access network functions include 5G UPF to handle 5G small cells and low-latency applications like CDNs.
4. OPEN BROADBAND TRIAL WITH DISTRIBUTED DEPLOYMENT MODEL
In the short term, it is mandatory that Open Broadband access is not only capable of providing innovation potential and long-term benefits, but it also needs to fully support legacy services and be interoperable with legacy network architecture and equipment.
To demonstrate the interoperability of the new Open Broadband architecture with the Telefónica legacy network, we performed an Open Broadband lab trial (see Fig. 6), following the distributed control plane deployment described in the former section.
For this trial, we used the Radisys Any-PON (G-PON or XGS-PON) OLT model RLT-1600X. It offers 16 small form-factor pluggable (SFP$+$) based Any-PON interfaces in 1RU height. An Any-PON interface means that each of these ports can support either a G-PON SFP or an XGS-PON SFP$+$. The OLT uses Aspen, the latest PON media access control (MAC) application specific integrated circuit (ASIC) from Broadcom, and the Qumran AX0 with 300 Gb/s switching capacity.
The OLT PON XGS-PON transceiver is an SFP$+$ module delivering ${+}{5.8}\;{\rm dBm}$ of optical power at 1577 nm nominal wavelength. Optical splitters and attenuators were used to emulate a PON and the optical power at the input of the XHGUs was around ${-}{19}\;{\rm dBm}$. These values are compliant with ITU-T G.9807.1 optical class Nominal2, which supports a maximum optical attenuation of 31 dB.
Two XHGUs from a third-party supplier were used in the trial and successful interoperability was achieved using a standard OMCI configuration. The SD-PON command line interface showing the operational status of one XHGU is shown in Fig. 7 (supplier is hidden for confidentiality reasons).
The OMCI provisioning model and the most relevant managed entities (MEs) are shown in Fig. 8, according to the ITU-T G.988 standard. The relationship between MEs is shown using arrows (dashed lines when the ME relationship is implicit). Squares with straight edges (PPTP VEIP/UNI, UNI-G) are automatically created by the XHGU, while the other MEs are created and configured by the SD-PON software. For diagram simplicity, a single priority queue (PQ) branch is included, and the T-CONT and PQ MEs for upstream transmission, as well as other optional managed entities, are not shown. To guarantee multivendor interoperability, vendor-specific managed entities have not been used.
The service configuration for the XHGU user network interface (UNI) is shown in Table 2 (for security reasons, the IP addresses shown are provided from a traffic generator and not from the production environment).
Table 2. Open Broadband Trial Service Configuration
The SD-PON software was run from within the OLT, and the OLT was connected to a ToR switch Edgecore—AS5912-54X for traffic aggregation and IP/multiprotocol label switching (MPLS) functions such as MPLS L2 virtual private networks. The ToR switch and the OLT are connected with a 10 Gb/s link, and the ToR switch was connected to the upstream network over a link aggregation group (LAG) of ${2} \times {10}\;{\rm Gb/s}$ links.
The tests predominantly used 10 Gb/s links even though 40/100 Gb/s links are also possible. The ToR switch was connected to Telefónica’s aggregation network that has subscriber service platform elements such as the BNG, voice over IP (VoIP), and video servers. The BNG provides access to the IP network through point-to-point protocol over Ethernet (PPPoE) sessions, and VoIP and video on demand (VoD) elements provided voice and video services. To connect the CPE to the service network elements over an L2 network, a subscriber’s L2 packets from the OLT are transported over MPLS pseudowires (PWs) from the ToR switch to the BNG, VoIP, and video servers (headend nodes).
At the XHGUs, a PPPoE client was used for Internet access and IP over Ethernet interfaces were used for video and voice services. The ToR switch configured with MPLS pseudowire in virtual private LAN service (VPLS) mode transported the stacked VLAN from the ToR switch to the headend nodes to establish the PPPoE session between the XHGU and the BNG for Internet service. Video and VoIP services were also delivered using stacked VLAN and transported in MPLS pseudowire interfaces configured in virtual private wire service (VPWS) mode in the ToR switch.
The Radisys management system (RMS), an EMS, managed the RLT-1600X OLTs with an in-band management setup. The in-band management was established through the ToR on the same link that connects the ToR and the OLT.
All Telefónica residential triple play services were successfully accessed over FTTH—high-speed Internet access, VoIP and IPTV with multicast TV and video on demand (VOD). Moreover, interoperability was achieved with Telefónica XHGUs as well as interoperability with third-party PON transceivers plugged into the OLT.
Maximum throughput between the OLT uplink and the XHGU local area network interfaces was tested versus the Ethernet frame packet size using a traffic generator and analyzer. The traffic emulation uses bidirectional IP traffic transmission between the OLT and the XHGU gateway over PPPoE. With the frame size of 1500 bytes, throughputs of about 8.6 Gb/s and 9.6 Gb/s were achieved, respectively, in the downstream and upstream directions. Downstream forward error correction (FEC) was enabled in the transmission layer of the XGS-PON interface.
We have measured the latency between the XHGU and the IP edge (BNG) in our trial scenario and obtained delays around 1 ms for both traditional OLTs and the Open Broadband equipment, thus the delay due to the virtualization layer in the Open Broadband architecture is negligible. The maximum power consumption in watts per PON interface for traditional PONs is obtained from the Code of Conduct on Energy Consumption of Broadband Equipment: Version 7.1 from the European Commission, and the maximum power consumption of the Open Broadband setup has been calculated from datasheet specifications for a configuration of 96 PON ports and one ToR switch.
Table 3 shows the round-trip time (RTT) delay and power consumption comparison between a traditional OLT and the Open Broadband equipment used in the trial.
Table 3. Comparison between a Traditional OLT and Open Broadband Equipment
5. FTTH NETWORK EVOLUTION WITH MULTI-PON TECHNOLOGY
Gigabit-capable PON standards were published nearly 20 years ago. In recent years, 10 Gb/s and 25 Gb/s capable standards [9–11] have been ratified with ITU-T 50 Gb/s PON currently in the process of standardization [12].
Traditional PON interfaces are usually composed of a SFP optical transceiver that support a single optical access technology. Additional OLTs with 10 Gb/s transceivers would need to be added in the central offices to accommodate 10 Gb/s access technologies in the same PON optical distribution network, and dedicated wavelength multiplexers would be necessary to combine legacy and 10 Gb/s optical signals in the same PON. This traditional approach would involve re-cabling within the central offices with additional space requirements for more OLTs, fiber cables, and multiplexers, along with an increase in power consumption.
As an alternative to traditional methods, multi-PON modules (MPM) have recently been considered in standard architectures [13] and new MPM cards have appeared on the market. In the MPM methodology, OLT PON interface cards can be configured with G-PON or XGS-PON optical transceivers, or they can be configured to support both G-PON and XGS-PON simultaneously on the same optical distribution network (ODN). SFP$+$ transceivers that combine both G-PON and XGS-PON optical channels with an integrated optical multiplexer are referred to as C-PON optics. These C-PON optics along with multi-PON OLTs, are key for network operators with existing G-PON deployments because they save space and cost as well as reduce operational complexity and ease the customer migration from G-PON to XGS-PON. They are also critical for new G-PON deployments because they can be easily upgraded to XGS-PON at any time in the future.
In the Telefónica Open Broadband initiative, whitebox OLTs are designed using multi-PON technology with interoperable third-party PON transceivers. G-PON SFPs are used for G-PON deployment, and XGS-PON can be seamlessly added on a per-PON-port basis simply by replacing a G-PON SFP module with a C-PON SFP$+$ module, enabling simultaneous access to G-PON and XGS-PON technologies in the same PON interface.
In our recent work [4], we shared the fundamentals for seamlessly transitioning from G-PON to XGS-PON systems using multi-PON technology. In this paper, we discuss the design challenges of C-PON transceivers and propose what we believe is a new transceiver design that supports legacy architecture and optimizes signal integrity.
The newer flexible OLTs come in two versions: Any-PON and multi-PON. In the Any-PON scenario, the OLT interface will support any of the single-channel PON technologies, primarily G-PON or XGS-PON, but not both. In contrast, multi-PON means that the OLT PON port will support those single-channel transceivers as well as both G-PON and XGS-PON simultaneously by employing the new combo-PON technology that combines the G-PON and XGS-PON wavelengths with an integrated wavelength multiplexer.
A. C-PON Optics Proposal: The Optical Transceiver as a Critical Path Forward
Although it is a small part, the OLT optical transceiver is a critical component in making the connection to residential and commercial clients. Historically, we have been able to incrementally evolve this interface, from G-PON and XGS-PON and then to multirate XGS-PON.
We are now moving to the next step with a dual-channel C-PON interface that supports simultaneous G-PON and XGS-PON with a single transceiver and a common optical link, and is able to work with multi-PON OLTs. As a community, we are ready to make that step to include C-PON in the set of available transceiver options. For that to happen, we need a common definition for the C-PON transceiver features and functions as well as the mechanical and electrical interface. We are not at this point today. To get there, we are proposing an open C-PON optic proposal.
Like previous PON interfaces, C-PON has been in migration over time. Initially, transceivers were based on one-at-a-time customer specifications for OLTs that only supported C-PON, but not single-technology G-PON or XGS-PON. These initial C-PON modules were not expected to support backward compatibility (i.e., be compatible with OLTs supporting single-channel transceivers). As a result, the pin definition was not aligned with earlier transceiver types. Several transceiver versions have been defined between specific customers and manufacturers under non-disclosure agreements (NDAs). These are unique to a product and are not compatible with each other.
The SFP-DD [14] transceiver form factor and pinout was considered since it is designed for two independent serial channels. However, it is only defined for Ethernet signals, has a much longer footprint, and is significantly more expensive. Although there is no common C-PON pinout, virtually all versions are based on the same SFP$+$ form factor and have a similar (although not identical) pinout to the other PON transceivers. There are many advantages to using that common form factor for C-PON, including reuse of the existing metal bodies and an internal opto-electric structure. Ultimately this reduces the transceiver cost, and the development time and effort, as well as the functional risks.
The objective now is to produce multi-PON OLTs that can fully support current legacy transceivers such as G-PON and XGS-PON. This is important to allow a gradual rollout of higher capacities while scaling investment over time. To be physically compatible, the new transceiver must have a similar form factor to the original SFP$+$.
Legacy transceivers were designed before C-PON, using a 20-pin double-sided card edge connector. They have a single differential bidirectional SerDes channel along with associated control and management signals. Additionally, the four differential signal pins have adjacent ground pins to ensure balanced high-speed signal return paths. C-PON transceivers have two channels and so some pins must be selected for the added SerDes and management signals.
The goal is to provide all the newer dual-channel benefits of C-PON while still providing the earlier socket pin functions and signaling of the legacy transceivers. This involves matching the position and functions of data and control signals, but also keeping the grounding that is associated with the high-speed differential transmitter (Tx) and receiver (Rx) signals. That is, the adjacent ground returns on the original G-PON and XGS-PON footprints must be maintained to guarantee that these signals will work reliably across all operating conditions.
Fig. 9. Critical differential signals and ground returns, two methods to support legacy G-PON and XGS-PON transceivers. (A) Ground returns do not match the differential signals and (B) Ground returns completely match the differential signals.
Some proprietary C-PON transceivers reuse the differential ground returns for other functions such as control signals, as shown in Fig. 9(A). However, removing local ground returns can create artifacts that distort differential wave shapes and add jitter in the high-speed Tx and Rx differential paths such as
- • introduction of common-mode noise,
- • crosstalk between control signals and differential Tx and Rx signals,
- • impedance discontinuities at transition points, and
- • reflections and attenuation.
To avoid such signal loss and jitter artifacts, our C-PON design provides balanced differential grounds, as shown in Fig. 9(B). Including all four ground returns provides the best signaling support [15] and completely matches the differential signals, power, and ground pin assignments of legacy G-PON and XGS-PON transceivers.
B. Clarifying the C-PON Interface
To have a common open C-PON transceiver footprint that is fully compatible with legacy G-PON/XGS-PON transceivers, we must meet the following requirements:
- • use the existing C-PON transceiver form factor and electro-optical internal design to eliminate redesign costs and delays;
- • allow an equipment footprint that can meet all the high-speed signaling requirements (differential signal grounds) of legacy transceivers;
- • follow the common existing C-PON transceiver form factor and optical and electrical design to eliminate redesign and delay; and
- • provide an open definition that is available without NDA to all manufacturers, equipment vendors, and operators.
Considering the former requirements, we propose an open transceiver definition that combines the features and functionality from the field of existing incompatible C-PON variants, as shown in Fig. 10. By keeping the form factor and internal optoelectronics intact and aligning the pin positions, all of the previously defined goals can be met.
Fig. 10. Physical pinout comparison between single-channel technology (left) and dual-channel C-PON transceivers (right).
Figure 10 shows the new whitebox OLT socket pinout that supports the new full-featured C-PON transceivers as well as legacy G-PON and XGS-PON (and other related types). Note that the legacy single-channel transceivers were defined with 20 pins, but the card edge has room for 22 pins. The migration to a combined two-channel adds many new signal requirements. Most current C-PON transceivers use all 22 pin positions to best support the additional SerDes channel and associated management signals. OLTs designed with the 22-pin socket can connect with any 20-pin or 22-pin transceiver. When legacy transceivers are installed in the new footprint, the two extra pins are simply not used. The goal with the new pinout is to add the new channel pins and functions while maintaining signal integrity on the legacy/primary channel.
The pin positions and functions are chosen to minimize changes from the legacy footprint while providing the additional differential interface and all associated channel controls. This was accomplished by the following:
- • using all 22 pins available on the transceiver connector when a C-PON transceiver is installed;
- • sharing some common functions between the two channels, such as a received signal strength indication (RSSI) trigger; and
- • moving some static or low-speed management functions from dedicated pins to the ${{\rm I}^2}{\rm C}$ interface. This uses existing standardized ${{\rm I}^2}{\rm C}/{\rm EEPROM}$ soft functions.
The ground signal labels (in green) highlight continuity of critical ground returns for legacy G-PON and XGS-PON transceivers when they are installed in a socket with the new pinout. Pin positions 2, 3, 7, and 11 (in red) highlight signal differences (on lower speed signals) between the legacy transceivers and the new pinout. OLTs with flexible PON interfaces must adjust the interface functions on these pins depending on the type of transceiver that is installed.
The support for migration from G-PON to XGS-PON to C-PON creates a more scalable, cost-effective optical ecosystem. The benefits of a common open C-PON transceiver definition are as follows:
- • a common configuration that can be used by manufacturers, equipment designers, and operators;
- • ensured compatibility between equipment and the available C-PON as well as legacy transceivers (G-PON/XGS-PON); and
- • clearly defined pin functions and I/O requirements.
These benefits are all contingent on competitively sourced, openly specified transceivers. C-PON transceivers will be key for G-PON network operators to deploy multi-PON OLTs, because it is expected that a major part of all G-PON ports will be migrated to C-PON in the future, as users or new services demand more capacity.
6. CONCLUSIONS AND FUTURE WORK
We have presented and demonstrated our Open Broadband initiative for the evolution of FTTH networks, based on hardware and software disaggregation, virtualization, and multi-PON technology. We have reported our successful Open Broadband lab trial with a distributed deployment model using Radisys Any-PON OLT equipment in XGS-PON operation and designed a new open C-PON module fully compatible with legacy G-PON/XGS-PON transceivers and optimized for signal integrity. We now intend to test the new architecture with commercial customers in the field using C-PON technology and prepare the migration to the centralized deployment model. We believe this open and disaggregated architecture provides a smooth migration path toward XGS-PON technology and has the potential to provide unprecedented flexibility to access network design and deployment, offer new services based on multi-access edge computing, and achieve the targets of the emerging 5G mobile network scenarios.
In the future, we should be able to demonstrate the easiness with which G-PON/XGS-PON whitebox OLTs can smoothly be upgraded by introducing the next generation (25 Gb/s or 50 Gb/s) whitebox OLTs and a simple software upgrade to the SD-PON software.
Acknowledgment
The authors wish to thank Eduardo Maltauro Nickel and Prathap Thammanna for their contributions and lab support during the Open Broadband trial. We also wish to thank Srikanth Yandapalli for his insights on SD-PON optimizations.
REFERENCES
1. Plume Cloud, “Active devices at home in the US,” Internet Usage Report, https://discover.plume.com/wfh-dashboard.
2. National Cable and Television Association (NCTA)—The Internet & Television Association, “COVID-19: how cable’s internet networks are performing,” 2021, https://www.ncta.com/COVIDdashboard.
3. “Cisco Annual Internet Report (2018–2023),” Cisco White Paper, 2020.
4. J. Montalvo, J. A. Torrijos, and D. Cortés, “New approaches in optical access networks to increase network flexibility and achieve 5G targets: an operator’s view,” in European Conference on Optical Communications (ECOC), December 2020, paper We1J-1.
5. “Cloud central office reference architectural framework,” Broadband Forum TR-384, Issue 1, 2018.
6. “Telefónica open access and edge computing,” White Paper, 2019.
7. Broadband Forum, “Open broadband—broadband access abstraction (OB-BAA),” https://obbaa.broadband-forum.org.
8. Open Networking Foundation, “Virtual OLT hardware abstraction,” https://wiki.opencord.org/display/CORD/VOLTHA.
9. “Standard IEEE 802.3av amendment 1: physical layer specifications and management parameters for 10 Gb/s passive optical networks” (IEEE, 2009).
10. “10-gigabit-capable symmetric passive optical network (XGS-PON),” ITU-T Recommendation G.9807.1, 2016.
11. “IEEE standard for Ethernet amendment 9: physical layer specifications and management parameters for 25 Gb/s and 50 Gb/s passive optical networks,” IEEE 802.3ca-2020, 2020.
12. D. Zhang, D. Liu, X. Wu, and D. Nesset, “Progress of ITU-T higher speed passive optical network (50G-PON) standardization,” J. Opt. Commun. Netw.12, D98–D108 (2020). [CrossRef]
13. “Enhancement band, Amendment 1, Appendix IV: Multi-PON module with integrated WDM. ITU-T,” ITU-T Recommendation G.984.5, 2018.
14. “SFP-DD Hardware Specification for SFP Double Density 2X Pluggable Transceiver, MSA, Revision 4.2,” 2020.
15. H. Johnson, High-Speed Signal Propagation (Pearson Education, 2003).
Julio Montalvo was born in Madrid, 1979, holds a M.Sc. degree in telecommunication engineering (2003) from the Technical University of Madrid and a Ph.D. degree in electrical, electronics and robotics engineering (2008) from Carlos III University of Madrid (awarded an Extraordinary Doctorate Prize). He is with Telefónica Research & Development, as a Fixed Access Expert in the Telefónica Global Systems and Networks CTIO. He is a Telefónica delegate at the Broadband Forum, ITU-T, and FSAN, and his current work involves optical fixed access technology planning, optical access evolution, and PON interoperability with devices. Dr. Montalvo is currently author of one granted patent (No. 103868 at Portuguese INPI, 2007) and more than 60 refereed publications in scientific journals (14 in JCR) and conferences. He is editor and author of chapter 2 of the book “Optical Transmission” (ISBN978-94-007-1766-4, Springer).
Jose Alfonso Torrijos has a computer science degree from the Polytechnic University of Madrid in 1992. In 1994, he joined Telefónica I$+$D working on different projects for Telefónica regarding IP, video, DTV, xDSL, and satellite technologies. In 2004 he joined the Broadband Access Technologies Division managing ARTES-1 and FP6 projects such as SATLIFE and SATSIX. In 2008, he started to work on fiber optic projects for Telefónica Spain regarding FTTH access and photonic networks. From 2010, he has been working on network efficiency and network evolution projects in the Fixed Access Network group at Telefónica’s Global CTIO office, participating as well in EU FP7 projects such as ACCORDANCE (OFDMA in PON systems) and COMBO (fixed-mobile network convergence) and following FSAN and BBF standardization activities.
Daniel Cortés was born in Alcalá de Henares, 1988, and holds a M.Sc. degree in telecommunication engineering (2011) from the Technical University of Madrid and a degree in economics (2019) from the National Distance Education University (UNED) of Spain. He has been in charge of Fixed Access Network Planning and Technology within Telefónica Global Systems & Networks since 2017. He is focused on PON technologies and copper evolution. His expertise includes network virtualization and FTTH network deployment.
Rajesh (Raj) Chundury is a telecom industry professional with 24 years of experience spanning wireline, wireless, IP transport, software-defined networking, network function virtualization, and open source technologies at fora such as ONF, ONAP, LFN, and O-RAN. He is the CTO/VP Customer Solutions at the Broadband Access Business unit of Radisys (enabling open telecom). Before his current role, Raj spent 23 years at Ericsson beginning in R&D and moving to customer-facing roles collaborating with wireless and wireline customers like AT&T, T-Mobile USA, and Sprint to design and deploy 3G, 4G, and 5G networks. He graduated with an M.S. in computer science from the University of Texas at Arlington.
Matt St. Peter is a platform hardware architect with over 25 years of experience with telecom and industrial product development. As part of Radisys Technology and Strategy his focus is on equipment design for broadband and RAN applications. This includes compute, switching, optical networks and synchronization in both central office and ruggedized outdoor installations. He has a B.E. in electrical engineering and a M.S. in computer science from the University of Delaware.
