Optical networks have become an indispensable infrastructure underpinning the digital economy and supporting the intensive data networking needs of industry, commerce, academic institutions, governments, and individuals worldwide^[1]. Since the amount of network traffic has increased dramatically, the capacity limit of conventional optical networks has been a concern. In addition, the non-transparent and non-flexible network nodes cause serious performance limits of the network system. To overcome these problems and realize a sustainable society, novel optical transmission and networking technologies capable of transporting ever-increasing data flows are necessary.

To extend the capacity limit of wavelength division multiplexing (WDM) optical networks, spatial division multiplexing (SDM) technologies^[2,3] and networking^[4] using multi-core fibers (MCFs)^[5] and various spatial modes^[6,7] have been recently studied and developed to increase the number of transmission channels in a single fiber. Indeed, few-mode (FM) MCFs with over 100 spatial channels have now been reported^[8,9], and single fiber capacity exceeding 2 Pb/s has been demonstrated^[10,11].

Homogeneous single-mode (SM)-MCFs, perhaps, offer the simplest migration path for adoption of high-capacity SDM technology in the near term, having been shown to support high spectral efficiency (SE) modulation formats and wide band operation without the complexity of high-order multiple input-multiple output (MIMO) based receivers. Furthermore, the relative uniformity of the homogeneous cores supports spatial super channels (SSCs) for shared transmitter hardware, digital signal processing (DSP) resources, and simplified switching^[12]. Additional system advantages can exploit the similar propagation characteristics of different SDM channels for self-homodyne detection^[13] or multi-dimensional modulation or coding^[14].

To improve the performance of optical networks, it is necessary to enhance the throughput of network nodes as well as to increase the transmission capacity. While the transmission capacity can be increased by various multiplexing methods (e.g., SDM) and multi-level modulation formats, the node throughput is generally limited due to the electronic processing capability in the node systems. Therefore, transparent optical switching technologies for various bit rates and formats without electrical processing can assist in high node throughput using little power. Moreover, in the near future, since various contents from small-data-size, low-quality content (e.g., E-mails, sensor data collection) to large-data-size, high-quality (e.g., high-definition video distribution, remote surgery) will be transported on networks, it is expected to employ suitable switching schemes for the property of contents to efficiently transmit them. Recently, the convergence of packet switch and circuit switch architectures on the control plane or data plane has received much attention for providing both best-effort and quality of service (QoS) guaranteed services on the same infrastructure^[15,16].

An optical packet and circuit integrated (OPCI) network, where optical packet switching (OPS) and optical circuit switching (OCS) technologies are strongly introduced, has been developed to increase the performance of node hardware^[17,18]. In OPCI networks, OPS links serve bandwidth-sharing and best-effort data transfer, while OCS links enable a fully occupied bandwidth and end-to-end QoS guaranteed data transport. Moreover, wavelength resources can be dynamically allocated to the OPS or OCS links according to traffic conditions.

In this Review, we review the recent progress on high-capacity systems using homogeneous SM-MCFs, and we describe our 2 Pb/s demonstration and the key fiber properties and the potential of system features to further the efficiency in high-capacity SDM transmission. Then, we present the recently developed OPCI node and show the performance.

Figure 1 shows fiber cross-sections and details of recent record capacity demonstrations. The first over 100 Tb/s transmission was achieved in 2011, using the first homogeneous trench-assisted seven-core fiber^[19], followed shortly by the 112 Tb/s transmission^[20], and extended a year later to 305 Tb/s in a homogeneous trench assisted 19-core fiber using wideband polarization division multiplexed (PDM) quadrature phase shift keyed (QPSK) modulation^[21]. By the end of 2012, the experimentally demonstrated transmission capacity already exceeded 1 Pb/s^[22] with 50 GHz spaced 222-channel WDM signals, each carrying 456 Gb/s PDM-32 quadrature amplitude modulation (QAM) single carrier frequency division multiplexing (FDM) in each core of a 52 km, 12-core homogeneous MCF. Shortly afterwards, 1.05 Pb/s transmission was achieved by using a combination of twelve SM cores carrying PDM-32QAM orthogonal FDM (OFDM) signals and two FM cores carrying PDM-QPSK in their LP01 and two LP11 modes^[23]. In 2015, two capacity demonstrations exceeding 2 Pb/s were reported. One set a record capacity of FM-MCFs with a 2.05 Pb/s transmission that was reported using an FM-MCF with 6 modes in each of the 19 cores^[10]. Each of the 114 spatial modes carried 360 C-band wavelength channels with duobinary data at 15 GBd, suggesting that if the SE and transmission band can be increased, such fibers, perhaps, offer the best possibility of attaining single fiber capacity towards and beyond 10 Pb/s.

At the same time, the suitability of homogeneous SM-MCFs for wideband, spectrally efficient modulation was demonstrated by the transmission of 22-core SSCs with a >10THz optical transmission bandwidth enabled by a wideband optical comb for transmission capacity of 2.15 Pb/s. The frequency comb source, custom designed by RAM Photonics, consisted of a narrow linewidth (5 kHz) seed laser modulated with a low noise 25 GHz oscillator with the resulting 25 GHz spaced comb spectrally broadened in a dispersion engineered fiber mixer^[24]. The 399 comb lines acted as 22 sub-channels of an SSC, each carrying PDM-64QAM at 24.5 GBd to transmit a net data rate of 2.15 Pb/s, assuming a 20% forward error correction (FEC) overhead and bit-error rates (BER) of <2.7×10−2, as shown in Fig. 2.

The homogeneous 22-core MCF was based on a new three-layer design with a two-pitch layout and a total cladding diameter of 260 μm, as shown in the top-right inset of Fig. 2. The 31.4 km span was spliced from five separately drawn sub-spans, giving rise to total link crosstalk (XT) including input/output couplers of −47 to −37.5dB at the comb seed wavelength of 1559 nm. As shown in Fig. 3, the spacing between rings means that the XT contribution to each non-center core is dominated by the two adjacent cores (black line) in the same ring. Hence, adopting a bi-directional transmission strategy^[25] to reduce the XT contribution from adjacent cores may reduce the total XT below −40dB/100km, below which little impact on the transmission distance of QAM formats of up to 64 QAM was observed^[26]. However, the impact of time-varying XT^[27] means an additional margin may be required. The low variation in propagation delays between SDM channels is expected to be advantageous for SSC transmission and resource sharing. For several core pairs of the 22-core fiber, this was measured to be <3ps over 24 h in lab conditions.

Figure 2 shows how such a transmitter and fiber could be integrated with such technologies, such as joint-core reception and SSC coding. The uncoded BER shows a strong wavelength dependence with sub-channel BERs varying by over 3 orders of magnitude. Such variation may be problematic for the selection of the outer FEC codes that are usually defined by a fixed overhead and the required input BER^[28]. The variation of pre-FEC BERs may be reduced by averaging the BER across channel groups. Indeed, Fig. 2 shows that the averaging over SSCs reduces the BER variation to just over 1 order of magnitude. However, this may be further reduced by using some of the total coding overhead to apply specific coded modulation, depending on the quality of each SSC^[29] to reach the target pre-FEC BER with the smallest possible overhead. Although SSCs may not be the optimum grouping strategy for such a system, they allow short optical codes of 100 s bits to condition the pre-FEC BERs of serially coded outer FEC using 1000 s to try and maximize the overall throughput.

To efficiently utilize the large transmission capacity, optical networking technologies are important. Optical switching techniques are expected to improve the switching capacity and flexibility. We have been developing an OPCI network to provide both a best-effort service and a QoS-guaranteed service by employing OPS and OCS, respectively. Users can select the desired services. Different wavelength resources are assigned for OPS and OCS links, and the amount of their wavelength resources are dynamically changed in accordance with the service usage conditions.

We have developed an OPCI node by introducing novel physical technologies^[17,18]. Figure 4 shows the configuration of the latest 2×2 OPCI node testbed for ring networks. The OPCI node for ring networks mainly consists of seven 10 Gb/s optical transport network (10G-OTN) transponders, two 100 Gb/s OTN (100G-OTN) transponders, a 100 Gb/s optical packet (100G-OP) transponder, two wavelength-selective switches (WSS) for add/drop functions, an OPS system, and some optical amplifiers. Figure 4 also shows the photograph of the OPCI node. The OPS system is composed of an electronic switch controller (SW-CONT), a broadcast-and-selective 2×2 electro-absorption (EA) switch subsystem, and burst-mode erbium-doped optical fiber amplifiers (EDFAs) with optical feedback. Wavelength resources are divided by waveband and allocated to OPS and OCS links. WSSs are used for combining or dividing OPS and OCS wavebands. Furthermore, two WSSs are used as an OCS system.

In OCS links, to send data on optical paths, a 10G-OTN transponder encapsulates 10 gigabit Ethernet (GbE) frames from a client network into the OTN format. In addition, a 100G-OTN transponder accommodates 10×10GbE frames or 100 GbE frames from the client side on the OTN format whose modulation scheme is dual-polarization (DP) QPSK. Because optical paths are established by control packets in advance, there is no need to read the IP destination address of incoming frames.

In OPS links, a 100G-OP transponder encapsulates an incoming 10 GbE frame from the client side into a 100 Gb/s multi-wavelength OP, which consists of ten 10 Gb/s on-off keying (OOK) optical payloads with different wavelengths and a destination optical label. The destination label is determined according to a mapping table between destination labels and the IP destination addresses of incoming 10 GbE frames. The 10 GbE frames, ranging from 64 to 9604 bytes, is directly encapsulated to an OP with no change of the IP address, MAC address, or VLAN-ID. The OP length varies according to the frame length. The 100G-OP transponder has multiple 10 GbE interfaces to accommodate many clients into OPCI networks. An SW-CONT reads the destination label and controls an EA switch subsystem to forward the OP to the correct output port according to a switching table in each input port. Also, control OPs for path signaling and wavelength resource control are exchanged via OPS links for the simplification of networks.

We demonstrated multi-format optical switching on the experimental setup of Fig. 4. We used 40 wavelength channels (named as λ1−λ40) ranging from 1531.90 to 1563.05 nm with 100 GHz grid spacing. The wavelengths were allocated to 10 Gb/s OOK optical paths from node 1 (λ11−λ17) and node 2 (λ18, λ19, λ32−λ36), 100 Gb/s DP-QPSK optical paths from node 1 (λ37, λ38) and node 2 (λ39, λ40), and 100 Gb/s OOK OPs (λ21−λ30).

First, we established fourteen 10 Gb/s OOK optical paths with a loop configuration. Network testers 1 and 5 transmitted/received 1518 byte 10 GbE frames to/from each 10G OTN transponder in nodes 1 and 2. Also, we established four 100 Gb/s DP-QPSK optical paths with a loop configuration. Testers 2 and 6 transmitted/received 64–1518 byte 10 GbE frames to/from a 100G OTN transponder with 100 GbE client interfaces in nodes 1 and 2. Testers 3 and 7 transmitted/received 1518 byte 10 GbE frames to/from a 100G-OTN transponder through ten 10 GbE client interfaces in nodes 1 and 2, respectively. Next, 100 Gb/s OOK OPs were launched to input port 2 of the EA switch in node 1. These OOK packets encapsulated 1,518 byte 10 GbE frames coming from ten 10 GbE client interfaces from tester 4. A switch controller read the route header on each OP to control the optical switch. OPs were switched to output port 1 in accordance with the routing table.

These OPs and paths were combined by a 9×1 WSS and transmitted over 50 km of SM fiber (SMF) to node 2. Figure 5 shows the spectral waveform of all optical signals in the output port of node 1. In node 2, an 8×1 WSS split the OPs and optical paths. In the EA switch of node 2, the OPS are switched to output port 1 and 2, respectively. A 100G-OP transponder received the packets and forwarded the decapsulated 10 GbE frames into network tester 8. The frame error rate (FER) for 10 GbE frames transmitted by 10 Gb/s OOK, 100 Gb/s DP-QPSK optical paths, and 100 Gb/s OOK OPs launched from node 1 is shown in Fig. 6. All measured FERs were much less than 1×10−4, which is regarded as high quality^[30].

This Review describes the high-capacity SDM transmission technologies to satisfy the ever-increasing traffic demands with a focus on recent 2 Pb/s demonstrations based on homogeneous, SM-MCF. This Review also shows the OP and circuit integrated network employing OPS and OCS to improve the switching capacity and flexibility in network nodes. The overall architecture of the OPCI node and the operation of the OPCI network testbed are reported.