Data Center Interconnects Research Articles

Joint equalization of CD and RSOP using a time–frequency domain Kalman filter structure in Kramers–Kronig system

Kramers–Kronig (KK) system is a direct detection (DD) optical short-reach interconnection system which is suitable for high capacity data center interconnect (DCI) applications. For this kind of short-reach high capacity polarization-division multiplexing (PDM) optical transmission system, the fiber channel impairments contain chromatic dispersion (CD) and rotation of state of polarization (RSOP) will lead to a severe signal distortion and need to be equalized. This paper proposed a time–frequency domain Kalman filter scheme for joint equalization of CD and RSOP simultaneously in KK system. The proposed scheme is capable of tracking the fast RSOP in time domain and equalizing CD in frequency domain using a window-split structure. The performance of the Kalman scheme is investigated in a 28GBaud PDM 16QAM DD system with KK receivers. The simulation results prove that the proposed Kalman scheme exhibits a good performance of RSOP tracking ability up to speed of 2.5 Mrad/s, with a large accumulated CD tolerance as large as 2040 ps/nm at a low OSNR penalty. The results also show the proposed scheme has a comparable computational complexity compared with a generally used scheme which includes a CD compensation, combined with polarization demultiplexing algorithm constant modulus algorithm (CMA) and multiple modulus algorithm (MMA).

Security-Aware Planning of Packet-Over-Optical Networks in Consideration of OTN Encryption

The fast development of cloud computing and Big Data applications has promoted virtualization technologies such as network function virtualization (NFV), which in turn dramatically increased the amount of sensitive data being transmitted over the optical networks for datacenter interconnections (DCIs). To ensure the physical-layer security in DCIs, people have developed optical transport network (OTN) encryption technologies, i.e., leveraging high-speed encryption cards (ECs) to encrypt OTN payload frames. Although experimental studies have confirmed the benefits of ECs in terms of line-speed processing, low latency, and small encryption overhead, the problem of how to utilize them to build a secure packet-over-optical network with high cost-effectiveness has not been explored yet. In this paper, we study how to realize cost-effective and security-aware multilayer planning in a packet-over-optical network that covers both trusted and untrusted zones, in consideration of OTN encryption. We first formulate an integer linear programming (ILP) model to minimize the total capital expenditure (CAPEX) of the multilayer planning, which includes the costs of OTN linecards (LCs), ECs, and bandwidth resources, and solve the optimization exactly. Then, we prove the <i xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">NP</i> -hardness of the multilayer planning, and to reduce the time complexity, we propose a column generation (CG) model and design a more time-efficient approximation algorithm based on it. Our simulation results confirm the performance and advantages of our CG-based proposal, i.e., it is much more time-efficient than solving the ILP directly, and outperform the existing heuristic in terms of total CAPEX and costs of used LCs and ECs.

Power-Efficient O-Band 40 Gbit/s PAM4 Transmitter Based on Dual-Drive Cascaded Carrier-Depletion and Carrier-Injection Silicon Mach-Zehnder Modulator With Binary Driving Electronics at CMOS Voltages

We present an O-band four-level pulse amplitude modulation (PAM4) transmitter based on a dual-drive silicon modulator with binary driving electronics at CMOS voltages, effectively eliminating the need for power-hungry digital-to-analog converters (DACs) and RF drivers. The fabricated dual-drive silicon modulator has cascaded p - n junction and p - i - n junction embedded in each arm of Mach-Zehnder interferometer, implementing high-speed data modulation and efficient bias control, respectively. Power dissipation of around 611 fJ/bit is obtained when synthesizing 46 Gb/s PAM4, showing high power efficiency. The fabricated silicon modulator exhibits a π-voltage-length product of 4.5 V·mm, and a 3-dB bandwidth of up to 27 GHz. Meanwhile, to implement a power-efficient PAM4 transceiver, the digital signal processing (DSP) at the receiver side is also discussed by balancing the performance and complexity. With a low-complexity DSP, negligible power penalty is observed after transmission over 38-km SSMF for both synthesized 20 Gb/s and 40 Gb/s PAM4. The presented PAM4 transmitter takes the advantages of both carrier-depletion and carrier-injection phase shifters for effective data modulation and bias control in PAM4 signalingand provides a power-efficient transmitter solution by eliminating the use of power-hungry DACs or RF drivers for datacenter interconnects.

Self restorable intra data center interconnect based on multimode fiber and free-space optics

We demonstrate a high data rate and low-cost OM4 multimode fiber (MMF) and free-space optics (FSO)-based self-restorable 2 × 100 Gbps per wavelength intra data center interconnect between two pairs of servers. This transmission system is based on two vertical cavity surface-emitting lasers that are directly modulated with PAM-4 data at the rate of 100 Gbps and multiplexed using mode division multiplexing of two linearly polarized modes. In case a fault occurs in the MMF-based primary path, the traffic is shifted to FSO-based protection path through a specially designed triggering and switching mechanism. Offline digital signal processing is used at the receiver to process the received PAM-4 signal for compensating the effects of channel impairments. The performance of the proposed architecture is evaluated at link lengths of 35 and 70 m under normal and faulty scenarios. The effect of varying the refractive index structure parameter of the log-normal channel model and the effect of mode coupling between linearly polarized (LP) modes on the bit-error-rate performance is also analyzed under different conditions. We observe that the performance of the FSO-based protection path is better than the OM4-based primary path at a forward error correction limit of 3.8 × 10 − 3 bit-error-rate.

Silicon modulator design using a system-oriented methodology for high-speed data center interconnect PAM-4 applications

A Silicon (Si) modulator and a design methodology using a system-oriented numerical approach are proposed. The main contributions of this work are to provide a Silicon modulator as an alternative solution for intra-data center interconnect (DCI) applications and a single figure-of-merit (FOM) for the modulator design, which takes as input the main device parameters (modulation efficiency (Vπ×L), optical loss, and electro-optical bandwidth (f3dB)) and provides a global optical system optimization. By using a numerical approach, the modulator physical parameters are analyzed according to system performance metrics such as the required receiver sensitivity to reach a target bit error rate (BER), and the required transmitter launch power. A capacitive Silicon device and a 4-level pulse amplitude modulation (PAM-4) system simulator is presented. The results show an efficient modulator (Vπ=3.5V) with high bandwidth (f3dB>38GHz), which is promising for data centers solutions, where power consumption is critical and high-speed is required.

Machine-learning-aided cognitive reconfiguration for flexible-bandwidth HPC and data center networks [Invited

This paper proposes a machine-learning (ML)-aided cognitive approach for effective bandwidth reconfiguration in optically interconnected datacenter/high-performance computing (HPC) systems. The proposed approach relies on a Hyper-X-like architecture augmented with flexible-bandwidth photonic interconnections at large scales using a hierarchical intra/inter-POD photonic switching layout. We first formulate the problem of the connectivity graph and routing scheme optimization as a mixed-integer linear programming model. A two-phase heuristic algorithm and a joint optimization approach are devised to solve the problem with low time complexity. Then, we propose an ML-based end-to-end performance estimator design to assist the network control plane with intelligent decision making for bandwidth reconfiguration. Numerical simulations using traffic distribution profiles extracted from HPC applications traces as well as random traffic matrices verify the accuracy performance of the ML design estimator ( &lt; 9 % error) and demonstrate up to 5 × throughput gain from the proposed approach compared with the baseline Hyper-X network using fixed all-to-all intra/inter-portable data center interconnects.

Variable Expanding Structure for Data Center Interconnection Networks

Data centers are fundamental facilities that support high-performance computing and large-scale data processing. To guarantee that a data center can provide excellent properties of expanding and routing, the interconnection network of a data center should be designed elaborately. Herein, we propose a novel structure for the interconnection network of data centers that can be expanded with a variable coefficient, also known as a variable expanding structure (VES). A VES is designed in a hierarchical manner and built iteratively. A VES can include hundreds of thousands and millions of servers with only a few layers. Meanwhile, a VES has an extremely short diameter, which implies better performance on routing between every pair of servers. Furthermore, we design an address space for the servers and switches in a VES. In addition, we propose a construction algorithm and routing algorithm associated with the address space. The results and analysis of simulations verify that the expanding rate of a VES depends on three factors:n,m, andkwhere thenis the number of ports on a switch, themis the expanding speed and thekis the number of layers. However, the factormyields the optimal effect. Hence, a VES can be designed with factormto achieve the expected expanding rate and server scale based on the initial planning objectives.

Real-Time Demonstration of Homodyne Coherent Bidirectional Transmission for Next-Generation Data Center Interconnects

We experimentally propose and demonstrate homodyne coherent detection in a short-distance data center interconnect with bidirectional fiber transmission. The transmitter in the proposed system sends the modulated signal and a continuous wave (CW) tone originating from the same laser diode (LD) over the 2 lanes of a duplex fiber to the remote receiver for coherent detection. Thus, this system allows the use of an uncooled LD with a large linewidth and reduces the complexity of digital signal processing (DSP) compared to that in a classical coherent system. A successful real-time demonstration with 600-Gb/s DP-64QAM using uncooled large-linewidth DFB lasers is conducted. Automatic stabilization against polarization fluctuations of the transmitted tone is achieved by the proposed polarization-tracking integrated coherent receiver in compact silicon photonics (SiP), tracking up to 300 rad/s without a performance penalty. The phase noise caused by mismatch and the link loss budget limitations are discussed and analyzed. This study shows that the proposed system is a potentially attractive solution for future 800G and 1.6T intra-data-center optical interconnects.

Optical Single Sideband Signal Reconstruction Based on Time-Domain Iteration

Due to its low cost, simple architecture and robustness to fiber dispersion, single sideband (SSB) transmission with direct detection (DD) system is an attractive solution for 80-km inter data center interconnects (DCIs). However, it will suffer performance penalty caused by the signal-to-signal beating interference (SSBI). Kramers-Kronig (KK) receiver has been extensively investigated for SSBI elimination by reconstructing the SSB signal. The nonlinear operations in KK algorithm require up-sampling to cope with spectral broadening, which results in high complexity for practical application. Optical signal phase retrieval method based on the minimum phase signal has also been investigated for SSB signal recovery, in which the SSB and DC-Value properties are iteratively imposed on the amplitude signal in frequency domain. In this paper, we propose a low complexity iterative algorithm for minimum phase signal recovery without up-sampling in time domain. Finite impulse response (FIR) filter is applied to iteratively generate the SSB signal and update the phase component. Based on the proposed scheme, the transmission of 30 GHz SSB 16-QAM discrete multitone (DMT) signal over 80 km single mode fiber (SMF) is successfully demonstrated with the bit error rate (BER) below the hard-decision forward error correction (HD-FEC) threshold of 3.8 × 10−3. The experimental results show that, the BER performance of KK scheme with up-sampling factor of 2, frequency-domain iterative scheme and our proposed scheme is almost the same. However, compared with the KK scheme, the proposed method can save the numbers of adders and multipliers by the factors of 29 and 7, while the factors are 5.5 and 4 comparing to the frequency-domain iteration scheme.

Outage Probability Due to Crosstalk from Multiple Interfering Cores in PAM4 Inter-Datacenter Connections

In this work, we propose to use four-level pulse amplitude modulation (PAM4) and multi-core fibers (MCFs) to support very high capacity datacenter interconnect (DCI) links. The limitations imposed by inter-core crosstalk (ICXT) on the performance of 112 Gb/s up to 80 km-long optically amplified PAM4 inter-DCI links with intensity-modulation and direct-detection and full chromatic dispersion compensation in the optical domain are analyzed through numerical simulation for high and low skew-symbol rate product (SSRP). With only one interfering core, we show that those PAM4 inter-DCI links achieve an outage probability (OP) of 10−4 with a maximum ICXT level of −13.9 dB for high SSRP and require an ICXT level reduction of about 8.1 dB to achieve the same OP for low SSRP. Due to using full dispersion compensation, for an OP of 10−4, the maximum acceptable ICXT level shows only a 1.4 dB variation with the MCF length increase from 10 km to 80 km. When considering the ICXT induced by several interfering cores, the maximum ICXT level per interfering core for an OP of 10−4 decreases around 3 dB when doubling the number of interfering cores. This conclusion holds for high and low SSRP regimes. For two interfering cores, we show that a single interfering core with low SSRP is enough to induce a severe reduction of the maximum acceptable ICXT level.

C band 112 Gb/s PAM4 signal transmission over 320 km with a quasi-linear double-side electro-absorption modulated laser (DS-EML)

The growing traffic demand in data center interconnections (DCIs) requires high-speed and low-cost transceivers that can flexibly adapt to different transmission distances up to 80 km or further. In this paper, based on the double-side electro-absorption modulated laser (DS-EML), we propose a transmitter with a fixed 90-degree phase rotation on one side of the DS-EML that can realize quasi-linear complex modulation. We verify the application of the quasi-linear DS-EML for a 112 Gb/s CD pre-compensation 4-level pulse amplitude modulation (PAM4) signal transmission over 320 km by simulation. Compared with the LiNbO3 in-phase quadrature (IQ) modulator and the dual-drive Mach-Zehnder modulator (DDMZM), our proposed transmitter has several obvious advantages, such as low cost, small footprint, ease of integration and reduced power consumption.

On-chip Ge, InGaAs, and colloidal quantum dot photodetectors: comparisons for application in silicon photonics

The past twenty years have seen explosive growth in silicon photonics technology. It has revolutionized numerous fields such high-speed optical interconnects in data centers. A photodetector (PD) is one of the key building blocks in silicon photonics, enabling on-chip light detection. Here a comprehensive study has been demonstrated in which three materials, germanium (Ge), indium gallium arsenide (InGaAs), and colloidal quantum dots (CQD), are compared for a PD integrated with a waveguide in silicon photonics. Comparisons are conducted by assuming InGaAs and CQD PDs have the same interface quality as mature Ge PD technology. With this premise, we intend to predict future InGaAs and CQD PD performances. Figures of merit such as dark current, responsivity, and RF bandwidth are compared using simulations. With the premise that epitaxial InGaAs on silicon is as of high quality as epi-Ge, results found that the InGaAs PD is advantageous over the Ge PD with higher-efficiency bandwidth product and lower dark current. CQD PD, on the other hand, is slow but has the lowest dark current, which is suitable for medium-speed applications where ultralow noise is required.

Time skew enabled vestigial sideband modulation for dispersion-tolerant direct-detection transmission.

Fiber dispersion and square-law detection-induced power fading is a fundamental obstacle in intensity modulation with direct detection links. In this Letter, we propose a hardware-efficient vestigial sideband (VSB) transmitter to suppress such an impairment. By introducing an appropriate time skew between the differential arms of the Mach-Zehnder modulator, a VSB signal can be generated based on a single digital-to-analog convertor without optical filtering. At the receiver, a Volterra equalizer is utilized to mitigate the inter-symbol interference. In the proof-of-concept experiment, 32 Gbaud 4/6/8-ary pulse amplitude modulation signals can be successfully transmitted over an 80 km standard single-mode fiber with bit-error rates below the KP4, 7% hard-decision and 20% soft-decision forward error correction thresholds, respectively. The proposed scheme provides a promising and low-cost solution for high-speed metro and data center interconnect applications.

POID: a passive all-optical inter-rack interconnect for data-centers

With the increase in demand for bandwidth-thirsty cloud-based services, bandwidth demands for rack-to-rack interconnect in data centers are increasing exponentially. In this paper, an optical interconnect architecture dubbed “POID” is proposed. The novel architecture is proposed using passive optical technology, wavelength division multiplexing, and innovative architectural concepts. The proposed architecture adequately supports the rack-to-rack traffic with latencies as low as few hundred nanoseconds while providing Tera bits per second throughput. The performance of the proposed architecture is evaluated through simulation studies and compared to other recent related data center interconnection networks from recent literature. The proposed POID architecture provides up to 10% improvement in throughput and up to about 82% reduction in latency compared to other comparative designs of data center interconnects using current optical fiber and switching technologies. The proposed PIOD incurs 22% power penalty compared to the existing best performing comparable designs.

128 GSa/s SiGe DAC Implementation Enabling 1.52 Tb/s Single Carrier Transmission

We report a new 128 GSa/s SiGe digital to analog converter (DAC) having 8 bit physical resolution and 24 GHz 3-dB bandwidth and present its architecture and the results on fundamental converter assessment. We further integrated the DAC into a coherent optical transmission testbed allowing increase of symbol rate up to 128 GBaud. Using probabilistic constellation shaped modulation formats at an entropy of 7.5 bits per symbol and polarization multiplex we successfully transmitted at net-bit rates up to 1.55 Tb/s for a single carrier in back to back while for a typical data center interconnect (DCI) distance of 80 km a net bit rate of 1.52 Tb/s has been achieved.

CMOS DAC Supported 1.1 Tb/s/λ DWDM Transmission at 9.8 bit/s/Hz Over DCI Distances

We report on a 16-nm CMOS DAC based transmitter optimization enabling bitrates up to 1.1 Tb/s. We describe the transmitter optimization and derive the maximization of the net bitrate. Based on back-to-back experiments we optimize the entropy of the probabilistically shaped QAM format, which we further apply for subsequent transmission experiments over data center interconnect distances. Finally, we successfully demonstrate DWDM transmission of 17 channels in a 112.5 GHz grid operated at 105 GBaud over 80 km up to 118 km achieving a net bitrate of 1.1 Tb/s and spectral efficiencies of 9.8 bit/s/Hz.

Silicon Photonic Crystal Modulators for High-Speed Transmission and Wavelength Division Multiplexing

For next-era optical interconnects in data centers, development of compact, energy-efficient, low-cost, and high-speed optical transceivers are required, for which high-performance external modulators in silicon photonics will be key components. We present a silicon photonic crystal waveguide slow light Mach-Zehnder modulator suitable for this purpose. The enhancement in the modulation efficiency via the slow light effect reduces the halfwave voltage-length product VπL, maintaining a wide working spectrum over 15 nm. The frequency response of the slow light modulator is constricted by an electrooptic phase mismatch between slow light and RF signals. In this study, this was dramatically improved by matching the phase using meanderline electrodes that delay RF signals. The cutoff frequency was experimentally evaluated to be 32-38 GHz. Using this device, we demonstrated high-speed modulation, including 64-Gbps on-off keying, 100-Gbps pulse amplitude modulation, and 50-Gbps/ch wavelength division multiplexing in 170-200-μm- long devices.

Versatile Externally Modulated Lasers Technology for Multiple Telecommunication Applications

This paper presents our technological approach for Externally Modulated Lasers (EMLs), based on Semi-Insulating Buried Heterostructure (SIBH) waveguide. We use Gas Source Molecular Beam Epitaxy (GSMBE) to grow the Phosphorus-based multi-quantum wells for both laser and modulator sections with butt-joint integration. The same GSMBE grows the p-doped InP claddings with low-diffusion Be dopant, leading to an accurate control of doping profiles, ensuring very steep modulator extinction curves. We present the main EML design rules and compromises, then apply them to different EMLs aiming at major telecom and datacom applications. After presenting characteristics of the standard 10 Gb/s C-band EML, we propose a 10 Gb/s EML at 1577 nm for next generation access networks, with a record high 10.5 dBm facet modulated power. Then we present high-speed EMLs up to 56 GBaud for datacenter interconnections, both in O- and C-band, aiming at very low peak-to-peak modulation voltage (<; 1.2 V), high facet modulated power (> 4.4 dBm), and compatible with uncooled operation (20 to 70 °C). These results confirm the efficiency and versatility of this technological platform for EMLs in a broad range of applications.

SiGe EAM-Based Transceivers for Datacenter Interconnects and Radio Over Fiber

Silicon photonics is a key-enabling technology leveraging decades of effort and infrastructure of the microelectronics CMOS industry resulting in high yield, low cost and potential high volume manufacturing. Furthermore, due to the high index contrast of the platform, very compact, high-complexity photonic integrated circuits can be devised. To benefit from these advantages, high-speed modulators should also be compatible with silicon technology. In this respect, SiGe electro-absorption modulators (EAM) are considered as a promising candidate since they are CMOS-compatible and offer high-speed, compact, low-loss and low-power modulation. In this paper, we discuss SiGe EAM-based transceivers for next-generation datacenter interconnects (DCI) and radio-over-fiber (RoF) fronthaul in next-generation cellular networks.

O-Band Subwavelength Grating Filters in a Monolithic Photonics Technology

The data communications industry has begun transitioning from electrical to optical interconnects in datacenters in order to overcome performance bottlenecks and meet consumer needs. To mitigate the costs associated with this change and achieve performance for 5G and beyond, it is crucial to explore advanced photonic devices that can enable high-bandwidth interconnects via wavelength-division multiplexing (WDM) in photonic integrated circuits. Subwavelength grating (SWG) filters have shown great promise for WDM applications. However, the small feature sizes necessary to implement these structures have prohibited them from penetrating into industrial applications. To explore the manufacturability and performance of SWG filters in an industrial setting, we fabricate and characterize the first O-band subwavelength grating filters in a monolithic photonics technology at GLOBALFOUNDRIES (GF). We demonstrate a low drop channel loss of −1.2 dB with a flat-top response, a high extinction ratio of −30 dB, a 3 dB channel width of 5 nm and single-source thermal tunability without shape distortion. This filter structure was designed using elements from the product design kit provided by GF and functions in a compact footprint of 0.002 mm2 with a minimum feature size of 150 nm.