Optical Data Center Research Articles

Synergistic Switch Control Enabled Optical Data Center Networks

Optical switching, which benefits from high bandwidth and low latency, promises to revolutionize data center networks that suffer from a bandwidth bottleneck and hierarchical structure. However, the lack of adaptable and fast optical switch control schemes, the lack of optical buffers for contention resolution, and the high cost of clock and data recovery (CDR) impede the practical deployment of optical switches in data centers. Networking communities have tried to address these challenges from different perspectives. However, the proposed solutions may be interdependent and thus introduce unexpected issues. In this article, a synergistic optical switch control mechanism is proposed to simultaneously solve the aforementioned challenges. The developed label control technique allows for controlling and configuring packet switching within 43.4 ns. The implementation of clock distribution achieves a fast CDR of 3.1 ns at low cost, and the design of an optical flow control protocol prevents packet loss caused by packet contention.

GNN-Based Hierarchical Deep Reinforcement Learning for NFV-Oriented Online Resource Orchestration in Elastic Optical DCIs

Network function virtualization (NFV) in elastic optical datacenter interconnections (EO-DCIs) enables flexible and timely deployment of network services. However, as the service provisioning of virtual network function service chains (vNF-SCs) in an EO-DCI needs to orchestrate the allocations of IT resources in datacenters (DCs) and spectrum resources on fiber links dynamically, it is a complex and challenging problem. In this work, we model the problem as a Markov decision process (MDP), and propose a hierarchical deep reinforcement learning (DRL) model based on graph neural network (GNN), namely, HRLOrch, to tackle it. To ensure its universality and scalability, we design the policy neural network (NN) in HRLOrch based on a GNN. As the GNN-based policy NN can operate on the graph-structured network state of an EO-DCI directly, it can adapt to an arbitrary EO-DCI topology without any structural changes. Then, through analysis, we find that the EO-DCI is a sparse reward environment if we want to train a DRL model to minimize the blocking probability of vNF-SCs in it directly. To address this issue, we design a hierarchical DRL with lower-level and upper-level models to improve the convergence performance of training. Specifically, we make the lower-level DRL optimize the provisioning scheme of each vNF-SC to minimize its resource usage, while the upper-level one coordinates the provisioning of all the active vNF-SCs to minimize the overall blocking probability. Hence, the lower-level and upper-level DRL models operate cooperatively in the training to optimize the dynamic provisioning of vNF-SCs. Our simulations demonstrate the universality and scalability of HRLOrch, and confirm that it can outperform the existing algorithms for vNF-SC provisioning in an EO-DCI.

Coflow scheduling and placement for packet-switched optical datacenter networks

Coflow scheduling and placement for packet-switched optical datacenter networks

On establishing and task scheduling of data-oriented vNF-SCs in an optical DCI

The development of network function virtualization (NFV) enables service providers to provision various network services with virtual network function service chains (vNF-SCs). However, most of the existing studies on the service provisioning of vNF-SCs address only the flow-oriented ones, while the provisioning schemes of data-oriented vNF-SCs, each of which needs to process and transfer bulk data through a series of vNFs before a preset deadline, have not been fully explored yet. Therefore, this paper studies how to jointly optimize the establishing and task scheduling of data-oriented vNF-SCs in an optical datacenter interconnection (DCI), such that the probability of scheduling data-oriented vNF-SCs to satisfy their deadlines can be maximized. To make the best use of the resources in the optical DCI for meeting each vNF-SC’s deadline, we leverage dynamic programming (DP) to propose two time-efficient algorithms with deadline-prioritized and conflict-aware approaches. Extensive simulations evaluate the performance of our proposed algorithms in different network scenarios, and confirm their effectiveness. Specifically, compared with a greedy-based benchmark, our DP-based algorithms can reduce the blocking probability of data-oriented vNF-SCs by up to two magnitudes and maintain similar service completion time.

A Fast Data Synchronization Technique in Optical Inter-Connected Data Centers

A Fast Data Synchronization Technique in Optical Inter-Connected Data Centers

Cerberus

The bandwidth and latency requirements of modern datacenter applications have led researchers to propose various topology designs using static, dynamic demand-oblivious (rotor), and/or dynamic demand-aware switches. However, given the diverse nature of datacenter traffic, there is little consensus about how these designs would fare against each other. In this work, we analyze the throughput of existing topology designs under different traffic patterns and study their unique advantages and potential costs in terms of bandwidth and latency ''tax''. To overcome the identified inefficiencies, we propose Cerberus, a unified, two-layer leaf-spine optical datacenter design with three topology types. Cerberus systematically matches different traffic patterns with their most suitable topology type: e.g., latency-sensitive flows are transmitted via a static topology, all-to-all traffic via a rotor topology, and elephant flows via a demand-aware topology. We show analytically and in simulations that Cerberus can improve throughput significantly compared to alternative approaches and operate datacenters at higher loads while being throughput-proportional.

Optical networks management and control: A review and recent challenges

In the last twenty years, optical networks have witnessed recurrent changes in their management and control architecture. In this paper, we present a historical timeline and a future perspective of the evolution of optical network management and control deployed for Wavelength Switched Optical Networks (WSON), Elastic Optical Networks (EON) and (multilayer) Data Center Networks.Early implementations of WSON envisaged a static and centralized provisioning approach supported by the Management Plane only. Gradually, the requirement of accommodating more network dynamicity in WSON, and later in EON, pushed the adoption of a distributed control, mostly supported by vendor-dependent implementations of the Generalized MultiProtocol Label Switching (GMPLS) protocol suite. The drawbacks of the fully distributed GMPLS-based control, such as resource contention, suboptimal resource usage, and complex computations (e.g., to account for physical layer constraints) showed the necessity to bring back some of the routing/provisioning functions to a centralized Path Computation Element (PCE) capable of accounting for e.g. physical impairments and interworking with GMPLS.The centralized control then gained its momentum and brought a radical change in network control, through the separation of data and control plane introduced by the paradigm of Software Defined Networking (SDN). Such an approach has been gradually extended to optical network control.The paper, eventually, presents the most advanced control techniques, namely the intent-based networking, the observe/decide/act state-based approach providing for autonomic optical network and the (closed-loop) zero-touch service management approach. Advanced traffic conditioning techniques are also detailed, namely the in-band telemetry and the exploitation of Programming Protocol-Independent Packet Processors (P4) language capabilities as well as solutions tailored for data center networks: all of them are still in a research stage and to be integrated within future optical network architectures.

Weight adjustable photonic synapse by nonlinear gain in a vertical cavity semiconductor optical amplifier

In this paper, we report a high-speed and tunable photonic synaptic element based on a vertical cavity semiconductor optical amplifier (VCSOA) operating with short (150 ps-long) and low-energy (μW peak power) light pulses. By exploiting nonlinear gain properties of VCSOAs when subject to external optical injection, our system permits full weight tunability of sub-ns input light pulses, just by varying the VCSOA's applied bias current. Not only is the VCSOA-based synapse able to adjust the strength of incoming optical pulses, but it can also provide gain (applied weight factors >1). Moreover, we show that this simple approach permits dynamical weight tuning at high-speed (ns rates) with up to an 11.6-bit precision. These results are realized with commercially sourced, inexpensive vertical cavity surface emitting lasers, operating at the key telecom wavelengths of 1300 and 1550 nm and hence making our approach compatible with optical network and data center technologies. This VCSOA-based system, therefore, offers a hardware friendly, low-energy, and high-speed solution for photonic synaptic links with high potential for use in future neuromorphic photonic systems.

Efficient Virtual Network Embedding of Cloud-Based Data Center Networks into Optical Networks

The demand for data center bandwidth has exploded due to the continuous development of cloud computing, causing the use of network resources close to saturation. Optical network has become an encouraging technology for many burgeoning networks and parallel/distributed computing applications because of its huge bandwidth. This article focuses on efficient embedding of data centers into optical networks, which aims to reduce complexity of the network topology by using the parallel transmission characteristics of optical fiber. We first present a novel virtual network embedding (VNE) mathematical model used for optical data center networks. Then we derive a priority of location VNE algorithm according to node proximity sensing and path comprehensive evaluation. Furthermore, we propose routing and wavelength assignment for DCNs into optical networks, and identify the lower bound of the required number of wavelengths. Extensive evaluations show that the proposed embedding algorithm can reduce the average waiting time of virtual network requests by 20 percent, increase the request acceptance rate and revenue-overhead ratio by 13 percent, as compared to the latest VNE algorithm.

Silicon photonic transceivers in the field of optical communication

Silicon photonics has developed rapidly in recent years, which has received widespread attention due to the fact that it can overcome the bandwidth bottleneck in optical communications. This paper mainly introduces the key applications of silicon photonics in the field of communication. Through a detailed description of optical transceiver modules in the coherent optical communication and data center, the advantages of silicon optical technology in the field of communication are introduced, such as high speed, low power consumption, high bandwidth and cost-effective. The problems of fabrication, packaging, light source integration and related devices in the current applications of silicon photonics are briefly analyzed. In the future, silicon photonics technology is expected to have better application prospects in long-distance transmission, big data communication and other communication applications.

Automatically reconfigurable optical data center network with dynamic bandwidth allocation

The rapid increasing traffic in data centers (DCs) puts tremendous pressure to the present multi-tier network architectures and electrical switching techniques. Switching traffic in the optical domain featuring ultra-high bandwidth, therefore, has been intensively investigated to build the high capacity data center networks (DCNs). To handle the variable traffic pattern in DCs, the network reconfigurability with adaptable optical bandwidth allocation is of key importance to flexibly assign the optical bandwidth. To this end, we propose and experimentally evaluate a software-defined networking enabled reconfigurable optical DCN with dynamic bandwidth allocation in this work, based on novel optical top of racks exploiting a wavelength selective switch. Experimental assessments show that the proposed solution can automatically reallocate the optical bandwidth in real-time to adapt the dynamic traffic pattern. Compared with the conventional optical DCN with static bandwidth provision, the end-to-end latency performance of the reconfigurable scheme with adaptable bandwidth allocation improves of 58.3% and the average packet loss decreases one order of magnitude. Moreover, the reconfigurable optical DCN features deterministic latency performance, with much lower time variations of packets delivery completion. Based on the experimental parameters, the simulation platform is also built to validate the good scalability of the proposed reconfigurable DCN. Numerical results illustrate the negligible performance degradation (11%) as the network scales from 2560 to 40 960 servers.

Preservation of shared backup path ink‐node content based elastic optical datacenter networks

AbstractEfficient path selection and content availability in the optical datacenter networks are essential for business continuity. To measure quantitatively the connectivity of content and give protection to diverse type of contents, the idea ofk‐node content connectivity is becoming popular. However, the assignment of dedicated spectrum for each end‐to‐end content path consumes too many resources. In order to enhance the resource efficiency, the spectrum can be shared between end‐to‐content paths of various requests. The spectrum sharing minimizes the backup spectrum requirement and has been a topic of research. In this study, an innovative model, namely, end‐to‐end content connectivity based quasi‐bipartite graphs (E‐QBG) for the possible path prediction and maximum matching by Hungarian path prediction (MM‐HPP) are implemented for the spectrum efficiency and the disaster resilient datacenter networks. The objective is to minimize the entire backup spectrum resource and operations path. The quasi‐bipartite graph is used for the end‐to‐end content backup route evaluation including spectrum sharing on various links, while the Hungarian algorithm is utilized for accurate maximum matching prediction between various end‐to‐end route sets. We introduce the concept of maximum dissortativity path based on the weight factor. The proposed E‐QBG and MM‐HPP reduces spectrum consumption which guarantees survivability against multi‐failure and natural disasters. This method leads to an approach for more effective utilization of the resources for end‐to‐end content connected elastic optical datacenter networks as compared with the existing techniques.

Cost-adaptive multi-class multicast service aggregation based on distributed sub-trees in elastic optical data center networks

Multicast provisioning is important for supporting applications such as data backup and cloud computing in elastic optical data center networks (EODCNs). As the traffic demand from these applications is quickly increasing, it is important to reduce the resource usage and implement cost-effective multicasting. By mapping multiple traffic flows into a single light-tree, the light-tree-based multicast service aggregation can reduce the number of required transceivers while accommodating many multicast requests. However, it may increase the spectrum usage, and it is not always feasible because some subscription-based or secure multicast services in the network should be sent only to authorized users. In this paper, we focus on the problem of routing, modulation level selection, spectrum, and transceiver assignment for multicast services with different security levels. We study potential performance gains resulting from applying a distributed sub-tree scheme to multi-class multicast service aggregation, realizing reduced overall cost. By formulating the mathematical models using an integer linear program (ILP), we obtain results for a small-scale network. We also propose a heuristic algorithm for large-scale networks, and results show that our proposed algorithm can be beneficial in terms of both spectrum and transceiver utilization in EODCNs.

Feasibility Study of Optical Wireless Technology in Data Center Network

We investigate the scalability and performance of a novel optical wireless data center network (OW-DCN) based on passive diffractive optics and fast tunable transmitters. The proposed OW-DCN architecture consists of two parallel inter- and intra- cluster networks to implement a flat communication between the top of racks within the DCN with at most two hops. The proposed DCN operation, the transmission performance and the scalability have been experimentally demonstrated and assessed for different modulation formats (NRZ-OOK and PAM4) with different sizes of DCN. A DCN with 8×8 racks shows 20Gbit/s OOK error free transmission with a power penalty of only 1dB compared to the B2B performance. For a 16×16 racks DCN, experiment indicates 16Gbit/s PAM4 transmission at FEC-limit of BER <; 2×10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-3</sup> . Moreover, scalability investigations of the proposed DCN architecture with optimized passive optics indicate that a 32×32 racks DCN with spectral band around 1 nm per link is feasible.

Optical data center switches design and analysis

Abstract Data centers are facing huge bottlenecks due to the emergence of high-speed data applications. To meet speed requirements, it is proposed to shift the current technologies toward optical communication-enabled data center (DC) devices. This paper discusses the recently proposed optical switch designs and proposes a novel switch design for optical DCs. The architectures proposed in the literature are compared in terms of notable parameters. Finally, proposed switch performance is measured in the network using Monte Carlo simulation and results are obtained in terms of packet loss probability, under a random traffic model while considering both nonperiodized and prioritized packets.

Neuromorphic encoding of image pixel data into rate-coded optical spike trains with a photonic VCSEL-neuron

Driven by the increasing significance of artificial intelligence, the field of neuromorphic (brain-inspired) photonics is attracting increasing interest, promising new, high-speed, and energy-efficient computing hardware for key applications in information processing and computer vision. Widely available photonic devices, such as vertical-cavity surface emitting lasers (VCSELs), offer highly desirable properties for photonic implementations of neuromorphic systems, such as high-speed and low energy operation, neuron-like dynamical responses, and ease of integration into chip-scale systems. Here, we experimentally demonstrate encoding of digital image data into continuous, rate-coded, up to GHz-speed optical spike trains with a VCSEL-based photonic spiking neuron. Moreover, our solution makes use of off-the-shelf fiber-optic components with operation at telecom wavelengths, therefore making the system compatible with current optical network and data center technologies. This VCSEL-based spiking encoder paves the way toward optical spike-based data processing and ultrafast neuromorphic vision systems.

A Fast Circuit Scheduling Scheme for Optical Data Center Networks

We propose a fast circuit scheduling algorithm capable of supporting optical switch with microsecond-scale switching delays. By simultaneously calculating the max-weighted matching for all optical switches, this scheduling algorithm releases the linear dependence of the time complexity upon the number of optical switches. Further by replacing the recursion and pointer operations with sorting and iteration, this algorithm can be easily accelerated by the hardware. The simulation results show that in large-scale network the fast circuit scheduling runs one or two orders of magnitude faster than most of existing scheduling algorithms, while only losing 4.08%~7.49% throughput comparing with the optimal solution. This improvement increases the circuit flexibility effectively and ultimately leads to 5.43%~79.63% and 30.45%~81.17% reduction in average and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$95th$ </tex-math></inline-formula> percentile flow completion time.

A novel framework for content connectivity through optical data centers

Abstract Nowadays, the data centers have become a significant physical infrastructure for the purpose of supporting various Internet applications like cloud services, entertainment, web search and social networking. The traffic among the data centers is growing rapidly accompanied by the services that are obtained. The data centers are connected via fiber channels to support long-haul networks and high data rate transmissions by utilizing various modulation techniques. However, it has some drawbacks such as increased delay, computational complexities, high wavelength consumption, link failures, etc. Recently, researchers are focusing on improving the survivability and wavelength usage efficiency in optical data center networks. In this work, a novel framework depending on the concept of content connectivity is proposed for optical data center networks. Here, a mixed integer linear programming (MILP) is utilized for transmitting the data through optical data centers. The main intention of this research is to improve the performance and wavelength efficiency in optical data center networks. The performance of the MILP approach is evaluated and compared with the existing integer linear programming (ILP) technique and found that this new approach provides better performance with higher wavelength efficiency and reduced wavelength consumption.

Passive Optical Interconnection Data Center Based on NRZ+Manchester Signal and Polarization Multiplexing

Passive Optical Interconnection Data Center Based on NRZ+Manchester Signal and Polarization Multiplexing

Experimental Assessments of SDN-Enabled Optical Polling Flow Control for Contention Resolution in Optical DCNs

Due to the lack of optical buffer, high packer loss caused by packet contention is one of the main challenges for the optical switching data centers (DCs). Flow control (FC) protocol employing electrical buffers at the top of racks (ToRs) and exploiting packet retransmission mechanism in case of contention has been extensively investigated to decrease the packet loss in optical DCs. However, the packet retransmission and the head-of-line (HOL) blocking in electrical buffers at substantial load traffic introduce extra latency and buffer-overflow that still cause packet loss. To overcome these issues, a novel contention resolution technique based on a software-defined networking (SDN) enabled optical polling flow control is proposed and experimentally assessed in this article. Experimental assessments show that the proposed contention resolution scheme achieves zero packet loss and 7.4 μs latency performance at the load of 0.4. In addition, we numerically modelled and investigated the scalability of the proposed contention resolution technique in a large scale DCN based on the experimental parameters. Results prove the excellent scalability performance of OPFC scheme, in which the packet loss increases from 2.1E-3 to 9.02E-3 and the average latency increases 5.17 μs at the load of 0.5 as the OPFC based network scales from 4 to 40960 servers.