Low-latency Processing Research Articles

FPGA-based adaptive space–time compression towards 5G MIMO fronthaul

Data compression-assisted radio-over-fiber (RoF) is a candidate solution to improve fronthaul (FH) bandwidth efficiency without losing cell-site simplicity for cloud/centralized radio access network (C-RAN) in fifth-generation (5G) mobile networks. One of the challenges in FH in 5G & beyond lies in MIMO fronthauling, where the FH bandwidth requirement scales (in spatial dimension) with number of antennas and thus cannot be resolved by conventional compression approaches tackling only time dimension. Also, it is desirable to implement FH compressor on real-time hardware to evaluate the implementation complexity, signal quality and processing latency. In this paper, we target 5G MIMO FH, and present field-programmable gate array (FPGA) based hardware design, characterization and evaluation of our proposed adaptive space–time FH compression technique. We experimentally demonstrate 8-antenna MIMO FH uplink enabled by the FPGA-based compressor. 48 Gb/s CPRI-equivalent rate encapsulating 1024QAM 5G new radio (NR)-like signals is transported using only 2.5 GBd optical PAM4, achieving <1.25% EVM with low processing latency. The FPGA demonstration indicates the feasibility of the space–time compression technique towards practice, which may be employed in not only FH network, but also other applications such as indoor MIMO distributed antenna system (DAS) or converged fiber-wireless transmission systems.

Massively Parallel Tree Search for High-Dimensional Sphere Decoders

The recent paradigm shift towards the transmission of large numbers of mutually interfering information streams, as in the case of aggressive spatial multiplexing, combined with requirements towards very low processing latency despite the frequency plateauing of traditional processors, initiates a need to revisit the fundamental maximum-likelihood (ML) and, consequently, the sphere-decoding (SD) detection problem. This work presents the design and VLSI architecture of MultiSphere; the first method to massively parallelize the tree search of large sphere decoders in a nearly-concurrent manner, without compromising their maximum-likelihood performance, and by keeping the overall processing complexity comparable to that of highly-optimized sequential sphere decoders. For a $10\times 10$10×10 MIMO spatially multiplexed system with 16-QAM modulation and 32 processing elements, our MultiSphere architecture can reduce latency by $29\times$29× against well-known sequential SDs, approaching the processing latency of linear detection methods, without compromising ML optimality. In MIMO multicarrier systems targeting exact ML decoding, MultiSphere achieves processing latency and hardware efficiency that are orders of magnitude improved compared to approaches employing one SD per subcarrier. In addition, for $16\!\times \!16$16×16 both “hard”- and “soft”-output MIMO systems, approximate MultiSphere versions are shown to achieve similar error rate performance with state-of-the art approximate SDs having akin parallelization properties, by using only one tenth of the processing elements, and to achieve up to approximately $9\!\times$9× increased energy efficiency.

A Reconfigurable Platform for Magnetic Resonance Data Acquisition and Processing.

Developments in magnetic resonance imaging (MRI) in the last decades show a trend towards a growing number of array coils and an increasing use of a wide variety of sensors. Associated cabling and safety issues have been addressed by moving data acquisition closer to the coil. However, with the increasing number of radio-frequency (RF) channels and trend towards higher acquisition duty-cycles, the data amount is growing, which poses challenges for throughput and data handling. As it is becoming a limitation, early compression and preprocessing is becoming ever more important. Additionally, sensors deliver diverse data, which require distinct and often low-latency processing for run-time updates of scanner operation. To address these challenges, we propose the transition to reconfigurable hardware with an application tailored assembly of interfaces and real-time processing resources. We present an integrated solution based on a system-on-chip (SoC), which offers sufficient throughput and hardware-based parallel processing power for very challenging applications. It is equipped with fiber-optical modules serving as versatile interfaces for modular systems with in-field operation. We demonstrate the utility of the platform on the example of concurrent imaging and field sensing with hardware-based coil compression and trajectory extraction. The preprocessed data are then used in expanded encoding model based image reconstruction of single-shot and segmented spirals as used in time-series and anatomical imaging respectively.

Low Resource Complexity R-peak Detection Based on Triangle Template Matching and Moving Average Filter.

A novel R-peak detection algorithm suitable for wearable electrocardiogram (ECG) devices is proposed with four objectives: robustness to noise, low latency processing, low resource complexity, and automatic tuning of parameters. The approach is a two-pronged algorithm comprising (1) triangle template matching to accentuate the slope information of the R-peaks and (2) a single moving average filter to define a dynamic threshold for peak detection. The proposed algorithm was validated on eight ECG public databases. The obtained results not only presented good accuracy, but also low resource complexity, all of which show great potential for detection R-peaks in ECG signals collected from wearable devices.

A 34-FPS 698-GOP/s/W Binarized Deep Neural Network-Based Natural Scene Text Interpretation Accelerator for Mobile Edge Computing

The scene text interpretation is a critical part of the natural scene interpretation. Currently, most of the existing work is based on high-end graphics processing units (GPUs) implementation, which is commonly used on the server side. However, in Internet of Things (IoT) application scenarios, the communication overhead from the edge device to the server is quite large, which sometimes even dominates the total processing time. Hence, the edge-computing oriented design is needed to solve this problem. In this paper, we present an architectural design and implementation of a natural scene text interpretation (NSTI) accelerator, which can classify and localize the text region on pixel-level efficiently in real-time on mobile devices. To target the real-time and low-latency processing, the binary convolutional encoder–decoder network is adopted as the core architecture to enable massive parallelism due to its binary feature. Massively parallelized computations and a highly pipelined data flow control enhance its latency and throughput performance. In addition, all the binarized intermediate results and parameters are stored on chip to eliminate the power consumption and latency overhead of the off-chip communication. The NSTI accelerator is implemented in a 40 nm CMOS technology, which can process scene text images (size of 128 × 32) at 34 fps and latency of 40 ms for pixelwise interpretation with the pixelwise classification accuracy over 90% on ICDAR-03 and ICDAR-13 dataset. The real energy-efficiency is 698 GOP/s/W and the peak energy-efficiency can get up to 7825 GOP/s/W. The proposed accelerator is 7 times more energy efficient than its optimized GPU-based implementation counterpart, while maintaining a real-time throughput with latency of 40 ms.

Stochastic Configuration Networks Based Adaptive Storage Replica Management for Power Big Data Processing

In the power industry, processing business big data from geographically distributed locations, such as online line-loss analysis, has emerged as an important application. How to achieve highly efficient big data storage to meet the requirements of low latency processing applications is quite challenging. In this paper, we propose a novel adaptive power storage replica management system, named PARMS, based on stochastic configuration networks (SCNs), in which the network traffic and the data center (DC) geodistribution are taken into consideration to improve data real-time processing. First, as a fast learning model with less computation burden and sound prediction performance, the SCN model is employed to estimate the traffic state of power data networks. Then, a series of data replica management algorithms is proposed to lower the effects of limited bandwidths and a fixed underlying infrastructure. Finally, the proposed PARMS is implemented using data-parallel computing frameworks (DCFs) for the power industry. Experiments are carried out in an electric power corporation of 230 million users, China Southern power grid, and the results show that our proposed solution can deal with power big data storage efficiently and the job completion times across geodistributed DCs are reduced by 12.19% on average.

High-Throughput FFT-SPA Decoder Implementation for Non-Binary LDPC Codes on x86 Multicore Processors

Low-Density Parity-Check (LDPC) codes are a well known Error Correction Code family used for instance, in wireless and satellite communication links. Error correction performance of LDPC codes was further enhanced by extending it to higher order Galois fields, giving rise hence to the so-called non-binary LDPC codes (NB-LDPC). Error correction performance improvement (CCSDS 2014) is the main reason behind the adoption by the Consultative Committee for Space Data Systems (CCSDS) of the NB-LDPC codes in the experimental specification for the future next generation uplinks (CCSDS 2014, 2015). The high error correction efficiency for short frames make NB-LDPC codes a good candidate for IoT applications. However, the performance gain comes at the expense of a high decoding computational complexity (CCSDS 2014; Conde-Canencia et al. 2009). In this paper, an x86 multicore NB-LDPC decoder implementation is provided. This decoder that implements the FFT-SPA algorithm provides a throughput improvement of about 1.3 × to 2.7 ×, a latency reduction of more than 95% and a power consumption halved in comparison with the most efficient works on GPU (Graphics Processing Unit) device. Indeed, an efficient memory mapping and computation optimizations on the x86 architecture enable to achieve a higher decoding throughput than the GPU-based in similar experimental setup. Consequently, the throughput efficiency, the low processing latency associated with a low power consumption makes this proposed multicore implementation practical and attractive for real time implementations of NB-LDPC decoders in future SDR or Cloud-RAN systems for CCSDS standard and IoT applications.

Spatial release from masking and sound localization using real-time sensorineural hearing loss and cochlear implant simulation

Simulations of sensorineural hearing loss (SNHL) and bilateral cochlear implantation (BCI) have been modeled successfully under static and non-real time conditions. This study performed two experiments testing the validity of a novel real-time SNHL/BCI simulation application for iOS using an in-ear binaural-recording headphone apparatus. The first experiment measured spatial release from masking (SRM) with normal hearing (NH), against headset apparatus simulations of NH, SNHL, and BCI using HINT sentences, speech shaped noise and forward masking. A one-sample t-test revealed significant differences between NH and simulated SNHL and BCI conditions showing reduced benefit from SRM. The second experiment employed noise bursts across nine frontal-plane loudspeakers and measured localization accuracy under NH and six frequency band BCI simulation conditions. Repeated measures two-way ANOVA and Cronbach’s Alpha suggested significantly reduced localization ability with BCI simulation. While further testing is needed, results here provide promising evidence that real-time binaural recording with low-latency processing and in-ear playback may be used to simulate SNHL and the BCI percept in NH listeners. The limitations and potential of this technology to expand the subject pool and expedite innovative testing are discussed.

Enhanced multicore–manycore interaction in high-performance video encoding

This paper presents an efficient cooperative interaction between multicore (CPU) and manycore (GPU) resources in the design of a high-performance video encoder. The proposed technique, applied to the well-established and highly optimized VP8 encoding format, can achieve a significant speed-up with respect to the mostly optimized software encoder (up to $$\times$$ 6), with minimum degradation of the visual quality and low processing latency. This result has been obtained through a highly optimized CPU–GPU interaction, the exploitation of specific GPU features, and a modified search algorithm specifically adapted to the GPU execution model. Several experimental results are reported and discussed, confirming the effectiveness of the proposed technique. The presented approach, though implemented for the VP8 standard, is of general interest, as it could be applied to any other video encoding scheme.

SDN Architecture for 6LoWPAN Wireless Sensor Networks.

Wireless sensor networks (WSN) are being increasingly used for data acquisition and control of remote devices. However, they present some constraints in critical and large-scale scenarios. The main limitations come from the nature of their components, such as lossy links, and devices with power supply limitations, poor processing power and limited memory. The main feature of software-defined networks (SDN) is the separation between the control plane and the data plane, making available a logically unified view of the topology in the controllers. In this way, it is possible to build network applications that take into account this unified view, which makes the SDN an alternative approach to solve the mentioned limitations. This paper presents the SD6WSN (software-defined 6LoWPAN wireless sensor network) architecture, developed to control the behavior of the data traffic in 6LoWPAN according to the SDN approach. It takes into account the specific characteristics of WSN devices, such as low data transfer rate, high latency, packet loss and low processing power, and takes advantage of the flexibility provided by flow-based forwarding, allowing the development of specific networking applications based on a unified view. We provide a detailed description of how we have implemented SD6WSN in the Contiki operating system. The new architecture is assessed in two experiments. The first considers a typical advanced metering infrastructure (AMI) network and measures the overhead of SD6WSN control messages in configurations involving different path lengths. The results indicate that the overhead introduced is not excessive, given the advantages that the SDN approach can bring. The second considers a grid-topology to evaluate the average latency of the peer-to-peer communication. It was observed that the average latency in the SD6WSN is considerably lower than that obtained with standard 6LoWPAN, showing the potential of the proposed approach.

Energy-Aware Mobile Edge Computing and Routing for Low-Latency Visual Data Processing

New paradigms such as Mobile Edge Computing (MEC) are becoming feasible for use in, e.g., real-time decision-making during disaster incident response to handle the data deluge occurring in the network edge. However, MEC deployments today lack flexible IoT device data handling such as handling user preferences for real-time versus energy-efficient processing. Moreover, MEC can also benefit from a policy-based edge routing to handle sustained performance levels with efficient energy consumption. In this paper, we study the potential of MEC to address application issues related to energy management on constrained IoT devices with limited power sources, while also providing low-latency processing of visual data being generated at high resolutions. Using a facial recognition application that is important in disaster incident response scenarios, we propose a novel “offload decision-making” algorithm that analyzes the tradeoffs in computing policies to offload visual data processing (i.e., to an edge cloud or a core cloud) at low-to-high workloads. This algorithm also analyzes the impact on energy consumption in the decision-making under different visual data consumption requirements (i.e., users with thick clients or thin clients). To address the processing-throughput versus energy-efficiency tradeoffs, we propose a “Sustainable Policy-based Intelligence-Driven Edge Routing” algorithm that uses machine learning within Mobile Ad hoc Networks. This algorithm is energy aware and improves the geographic routing baseline performance (i.e., minimizes impact of local minima) for throughput performance sustainability, while also enabling flexible policy specification. We evaluate our proposed algorithms by conducting experiments on a realistic edge and core cloud testbed in the GENI Cloud infrastructure, and recreate disaster scenes of tornado damages within simulations. Our empirical results show how MEC can provide flexibility to users who desire energy conservation over low latency or vice versa in the visual data processing with a facial recognition application. In addition, our simulation results show that our routing approach outperforms existing solutions under diverse user preferences, node mobility, and severe node failure conditions.

Digital Signal Processing for High-Speed Fiber-Wireless Convergence [Invited

We review recent advances in the field of fiber-wireless convergence to better support cloud radio access networks via high-speed low-latency mobile fronthaul between remote radio units and baseband units. We also present how digital signal processing (DSP) may be used to support both common public radio interfaces (CPRI) in a point-to-point architecture and Ethernet-based CPRI in a point-to-multipoint architecture to achieve high bandwidth efficiency and low processing latency. The use of DSP-enabled next-generation passive optical networks (PON), such as 50 Gb/s PON, to support mobile fronthaul in a resource-efficient manner will also be discussed.

Real-time dispersion interferometry for density feedback in fusion devices

Interferometry as one of the most common core fusion diagnostics has traditionally suffered from incomplete vibration compensation. Dispersion interferometry promises a more complete compensation of vibrations. For this reason it is being employed in an increasing number of experiments. However, thus far none of them have shown reliable real-time low-latency processing of dispersion interferometry data. Nonetheless this is a necessity for most machines when trying to do density feedback control, most notably in long discharges like the ones planned at the W7-X stellarator and ITER. In this paper we report the development of a new phase extraction method specifically developed for real-time evaluation using field programmable gate arrays (FPGA). It has been shown to operate reliably during the operation phase OP1.2a at W7-X and is now routinely being used by the W7-X density feedback system up to very high densities above 1.4×1020 m-2 without showing 2π-wraps and exhibits increased wrap stability by double-data-rate sampling. A rigorous error analysis has been conducted shedding insights into the signal composition of a dispersion interferometer have been gained. This includes the environmental effects, most notably air humidity, on the phase measurement and the correction thereof.

BES: Differentially private event aggregation for large-scale IoT-based systems

The emergence of Internet of Things (IoT) offers many advantages, but it also raises significant challenges with respect to efficient and distributed processing of large data and also privacy concerns related to large data disclosure.We investigate the above problems from a system-perspective and study how differential privacy can be used to complement other privacy-enhancing technologies to allow for controlled large data disclosure. We present a streaming-based framework, Bes, where we leverage the often distributed nature of typical IoT systems for efficient computation of differentially private aggregates. We also propose methods to limit the noise that is commonly introduced for differential privacy in real-world applications, by bounding the outliers based on (differentially private) parameters of the actual system at hand or data from other similar systems.We also provide a thorough evaluation based on a fully implemented Bes prototype using real-world data from of a concrete IoT system, namely an Advanced Metering Infrastructure (AMI). We show how a large number of events can be aggregated in a private fashion with low processing latency, even when the processing is made by a single-board device, with similar capabilities to the devices deployed in AMIs. Moreover, by implementing a de-pseudonymization attack known from the literature, we also show the strong complementary protection offered by Bes’ differentially private aggregation, compared to other privacy-enhancing technologies.

Viper: A module for communication-layer determinism and scaling in low-latency stream processing

Stream Processing Engines (SPEs) process continuous streams of data and produce results in a real-time fashion, typically through one-at-a-time tuple analysis. In Fog architectures, the limited resources of the edge devices, enabling close-to-the-source scalable analysis, demand for computationally- and energy-efficient SPEs. When looking into the vital SPE processing properties required from applications, determinism, which ensures consistent results independently of the way the analysis is parallelized, has a strong position besides scalability in throughput and low processing latency. SPEs scale in throughput and latency by relying on shared-nothing parallelism, deploying multiple copies of each operator to which tuples are distributed based on its semantics. The coordination of the asynchronous analysis of parallel operators required to enforce determinism is then carried out by additional dedicated sorting operators. To prevent this costly coordination from becoming a bottleneck, we introduce the Viper communication module, which can be integrated in the SPE communication layer and boost the coordination of the parallel threads analyzing the data. Using Apache Storm and data extracted from the Linear Road benchmark and a real-world smart grid system, we show benefits in the throughput, latency and energy efficiency coming from the utilization of the Viper module.

Cooperative Computing System for Heavy-Computation and Low-Latency Processing in Wireless Sensor Networks.

Over the past decades, hardware and software technologies for wireless sensor networks (WSNs) have significantly progressed, and WSNs are widely used in various areas including Internet of Things (IoT). In general, existing WSNs are mainly used for applications that require delay-tolerance and low-computation due to the poor resources of traditional sensor nodes in WSNs. However, compared to the traditional sensor nodes, today’s devices for WSNs have more powerful resource. Thus, sensor nodes these days not only conduct sensing and transmitting data to servers but also are able to process many operations, so more diverse applications can be applied to WSNs. Especially, many applications using audio data have been proposed because audio is one of the most widely used data types, and many mobile devices already have a built-in microphone. However, many of the applications have a requirement that heavy-operations should be done by a tight deadline, so it is difficult for a single node in WSNs to run relatively heavy applications by itself. In this paper, to overcome this limitation of WSNs, we propose a new emerging system, HeaLow, a cooperative computing system for heavy-computation and low-latency processing in WSNs. We designed HeaLow and carried out the practical implementation on real devices. We confirmed the effectiveness of HeaLow through various experiments using the real devices and simulations. Using HeaLow, nodes in WSNs are able to perform heavy-computation processes while satisfying a completion time requirement.

Low-complexity time synchronisation for 3-dimensional underwater communications environment using estimated instantaneous node velocity

Underwater time synchronisation is required to use the latest communication technologies. Because propagation time should be estimated to improve time synchronisation accuracy, node mobility is an important factor. In the subaqueous environment, furthermore, it is impossible to exclude the influence of the node movement occurred by environmental factors and underwater nodes could have self-mobility. Conventional time synchronisations have considered node mobility but have an unrealistic assumption of node movement and high computational complexity. In this paper, however, our novel method, Low-complexity time synchronisation for 3-dimensional underwater communications environment using estimated instantaneous node velocity based (V-Sync), shows high accuracy with low complexity and short processing latency. In this paper, it is supposed that mobile nodes could know their instantaneous speeds when they receive the synchronising messages. Time synchronisation errors, skew and offset, are estimated by the instantaneous speeds. The performance and analysis of V-Sync in the underwater mobile environment is demonstrated by simulation results.

Guest Editorial First Edition of Series On Network Softwarization and Enablers

Mobile and fixed networks, along with service delivery platforms, must answer to the constantly growing highly broadband accesses, massive data usage, and very low latency services. They should ensure flexible and faster service revenues, and reduce expenditure and operational costs of next generation (5G and beyond) networks. Also, the generation and expansion of Internet of Things, Internet of Services, and the blooming of Network integrated Cloud Services have pushed the horizon of new computing paradigms and service-aware and multi-purpose networks characterized by proximity, low latency, high bandwidth, and real-time processing.

Flexible data structures for dynamic virtual auditory scenes

Virtual environments and their contents are dynamically changing, but also need to respond to the user immediately. While managing a dynamic scene is a common and well-understood problem for visual rendering, additional challenges exist for the high-quality audio rendering of such scenes. Audio rendering differs in a key aspect: Sound waves propagate substantially slower than light. For the acoustics in scenes of large dimensions, it is not sufficient to regard just the state at the current time. The sound propagation times become so significant (perceptible) that the past of the objects matter, making a time history of the scene necessary. Particularly the conjunction of multithreading and low-latency audio processing makes the description of the virtual acoustic scene a problem on its own. This paper presents a novel solution to this acoustic-related problem. We discuss the challenges of realizing a real-time auralization on modern (non-real-time) operating systems and state the main requirements of the data structure. A hierarchical state-based data structure with time history is presented, which not only fulfills the requirements for outdoor auralizations but also has key advantages for indoor simulations—such as room acoustics. A key feature is the integral support of atomic scene modifications, allowing several modifications to be performed at the same time. The presented concept is very modular and beneficial for a wide range of applications.

Task Scheduling for Edge Computing with Agile VNFs On‐Demand Service Model toward 5G and Beyond

The demand for satisfying service requests, effectively allocating computing resources, and providing service on‐demand application continuously increases along with the rapid development of the Internet. Edge computing is used to satisfy the low latency, network connection, and local data processing requirements and to alleviate the workload in the cloud. This paper proposes a gateway‐based edge computing service model to reduce the latency of data transmission and the network bandwidth from and to the cloud. An on‐demand computing resource allocation can be achieved by adjusting the task schedule of the edge gateway via the lightweight virtualization technology, Docker. The edge gateway can also process the service requests in the local network. The proposed edge computing service model not only eliminates the computation burden of the traditional cloud service model but also improves the operation efficiency of the edge computing nodes. This model can also be used for various innovation applications in the cloud‐edge computing environment for 5G and beyond.