Latency Performance Research Articles

Thermal Performance Enhancement With DRX in 5G Millimeter Wave Communication System

The evolution of the communication and computation systems enables the user equipment (UE) to handle tremendous transmission data. However, the high-speed data processing also makes UEs release heat and might burn the chips inside the devices. The thermal issue would be more critical in millimeter-wave communications. The massive antenna arrays and the radio frequency modules not only drain the UE battery but also heat the devices. 3GPP also identified the thermal issue and suppressed heat generation by temporarily reducing UE capability in Release 15. In this work, instead of reducing the UE capability, we propose to apply the beam-aware Discontinuous Reception (DRX) mechanism to manage the power consumption and temperature of UEs simultaneously. We are the first to analyze the temperature for UEs with DRX configured. A semi-Markov model is provided, and we employ it to estimate the sleep ratio, packet delay, and steady temperature. We use a simulation program to verify the proposed analytical model. When comparing the beam-aware DRX with the baseline 5G NR DRX operation, we find that the beam-aware scheme reduces the steady temperature from $38.2^\circ C$ to $26.7^\circ C$ . The results show that Beam-Aware DRX could solve the thermal issue without sacrificing much performance of packet delivery latency.

Slot Reallocation and Rejection for Collision Avoidance in Autonomous TSCH Networks

The IEEE 802.15.4-2015 TSCH (Time Slotted Channel Hopping) standard provides high-reliability, deterministic delay, and energy efficiency with the channel hopping and the TDMA. The standard does not specify the scheduling method, which is an essential part of the solution. One of the representative studies in TSCH scheduling is Orchestra. It is an autonomous scheduling method with minimal overhead that does not require a negotiation process between nodes. However, with high traffic load, performance of the Orchestra is greatly degraded because of its fixed schedule. If a node has children, continuous collisions occur in the slot of the node. We propose a novel algorithm called SRCA (Slot Reallocation for Collision Avoidance), as a solution for this problem. The SRCA autonomously updates the schedule. When a child requests a slot reallocation, the parent allocates a new slot with a low probability of collision. Another algorithm called SJCA (Slot reJection for Collision Avoidance) is further proposed in this work. It improves the SRCA to operate in an environment with interference where a sending node can disturb a transmission from a hidden node. In the SJCA, a node determines whether a collision due to interference will occur in the slot newly allocated by the parent. If so, it rejects the slot and requests again. We compare the performance of the SRCA and the SJCA against the Orchestra and e-TSCH-Orch through simulations. It is confirmed that the SJCA is robust to interference. In most cases, the SJCA shows better performance of PAR (Packet Acknowledgement Ratio), PLR (Packet Loss Rate), and latency than others. It is also verified that the SJCA provides high-reliability and low latency compared to existing technologies.

SLA-Aware Multi-Criteria Data Placement in Cloud Storage Systems

This paper proposes a multi-criteria data placement mechanism for typical cloud storage systems. The primary goal of this study is to place the clients’ files in the storage cluster by taking into account the principles of service level agreements and the levels of user support services. For this purpose, a multi-criteria scoring model is defined based on a trade-off between three criteria that can be crucial from the clients’ viewpoint: data access latency, data privacy, and computational performance. The second goal of the proposed mechanism is to improve the fairness of the storage allocation mechanism for various kinds of files compared to the conventional mechanisms. In this case, the cloud service provider gives each data storage node a chance to become the host of an already received file using a probabilistic ranking model. The experimental results demonstrate that the proposed data placement mechanism increases the robustness of cloud storage architecture under different file, network, and storage configurations compared to the existing data placement mechanisms.

SGedge: Stochastic Geometry-Based Model for Multi-Access Edge Computing in Wireless Sensor Networks

Multi-access edge computing (MEC) is used to provide low-latency computing services for users and offload computation-intensive tasks from users. The MEC network has two parts, a server and wireless sensor access networks, with the latter all independently connected to the server. In this study, the limited latency performance of a MEC network—namely, the communication and computation latency under coverage of the sensor access networks, as well as the stability of the server—is investigated. Moreover, referring to network constraints, the scaling laws are determined, which are for both communication and computation latency regarding network loads and resource parameters. The deduced scaling laws quantify the trade-offs between computing latency, network coverage, and stability.

Analysis of NOMA-Based Retransmission Schemes for Factory Automation Applications

New use cases and applications in factory automation scenarios impose demanding requirements for traditional industrial communications. In particular, latency and reliability are considered as some of the most representative Key Performance Indicators (KPI) that limit the technological choices addressing wireless communications. Indeed, there is a considerable research effort ongoing in the area of wireless systems, not only from academia, but also from companies, towards novel solutions that fit Industry 4.0 KPIs. A major limitation for traditional wireless architectures is related to the harsh nature of the industrial propagation channel. Accordingly, this paper addresses these challenges by studying the reliability and latency performance of the joint use of different retransmission schemes in combination with Non-Orthogonal Multiple Access (NOMA) techniques. Two general retransmission schemes have been tested: time-based and spatial diversity-based retransmissions. An adaptive injection level NOMA solution has been combined with the retransmission schemes to improve the reliability of critical information. In all cases, a particular set of simulations has been carried out varying the main parameters, such as modulation, code rate and the injection level. Moreover, the impact of the number of transmitters in relation to the communication reliability has been analyzed. Results show that spatial diversity-based retransmissions overcome considerably the reliability obtained with time-domain retransmissions while maintaining assumable latency rates.

Optical service unit (OSU)-based next generation optical transport network (NG OTN) technology and verification

With the development of transmission technologies, optical transport network (OTN) has evolved from a digital encapsulation technology to a network technology that supports multi-service transport. As the OTN gradually expands to the edge of the metropolitan area network (MAN) and enterprise users have higher quality requirements on bandwidth and service security, more and more private line services are provided via OTN network. Therefore, the OTN container is transformed from an optical data unit (ODU) to an optical service unit (OSU). The OTN industry has studied several OSU technologies to fit the development trends of NG OTN, aiming at transporting 2 Mbit/s to 100 Gbit/s constant bit rate (CBR) services and Ethernet services with high efficiency and more flexibilities, as well as reducing OTN network deployment and operation and maintenance (O&M) costs. In this work, several key technologies in NG OTN are introduced including NG OTN technical architecture, OSU frame format, client mapping and OSU multiplexing. Based on these technologies, we (China Mobile) took the lead in conducting the industry's first NG OTN test in Qingdao, which verified the service transport capability and lower latency performance of OSUs and promoted the maturity of the NG OTN industry chain.

ETM: Effective Tuning Method Based on Multi-Objective and Knowledge Transfer in Image Recognition

With the widespread application of machine learning and deep learning, image recognition has been continuously developed. However, there are still huge challenges in the use of machine learning and deep learning. The tuning processes of algorithms are critical and challenging for their performance. Although there have been many previous works to improve the final accuracy of the recognition algorithms through tuning, these works cannot consider some indicators that are also very important in the actual environment (such as latency, central processing unit (cpu) utilization) in the tuning. In this paper, we propose an effective tuning method based on multi-objective and knowledge transfer, which is solved the above limitations in the image recognition. Specifically, we first use an agent to automatically tune the recognition algorithms, and combine the prediction accuracy and the running latency of each episode as a multi-objective reward signal to guide the update of the internal parameters of the agent. In this way, the agent can continuously select the better algorithm configuration to improve prediction performance. In addition, we improve the efficiency of the above tuning process by transferring knowledge. To do that, we can learn the meta parameters from other small-scale tasks to initialize the agent. In the experiments, we apply the proposed method to tune the eXtreme Gradient Boosting and random forest on 57 image recognition tasks and convolutional neural network on 2 tasks. The experimental results verify that the proposed method achieves average accuracy rankings of 1.92, 1.42 and 1.71 on three algorithms to be optimized, respectively. Especially in terms of latency performance, the proposed method performs best on all the tasks (57 data sets) on the three algorithms to be optimized. In addition, we verify the various components of the proposed method through ablation experiments.

Fair and efficient DWBA algorithm based on SLA differentiated polling interleaved scheduling for NG-EPON

This study, based on the Next Generation Ethernet Passive Optical Network (NG-EPON) facing more complex and changeable network, develops a flexible and efficient resource allocation algorithm. First, a differentiated polling interleaved scheduling (DPIS) framework is designed based on dynamic awareness, significantly reducing idle time slots, improving channel upstream utilization, and performing service-level-agreement (SLA) support and Optical Network Unit (ONU) overall scheduling. Subsequently, a dynamic wavelength and bandwidth allocation (DWBA) scheme is presented, which allocates fair bandwidth between and in subcycles and ensures the timely transmission of large traffic ONUs. As revealed from the simulation results, the developed DPIS-DWBA algorithm exhibits high utilization and low latency performance of the online scheduling algorithm; it can also achieve the offline scheduling to support SLA and fairness and offers more effective delay differentiation and bandwidth fair service.

Privacy‐Aware Online Task Offloading for Mobile‐Edge Computing

Mobile edge computing (MEC) has been envisaged as one of the most promising technologies in the fifth generation (5G) mobile networks. It allows mobile devices to offload their computation‐demanding and latency‐critical tasks to the resource‐rich MEC servers. Accordingly, MEC can significantly improve the latency performance and reduce energy consumption for mobile devices. Nonetheless, privacy leakage may occur during the task offloading process. Most existing works ignored these issues or just investigated the system‐level solution for MEC. Privacy‐aware and user‐level task offloading optimization problems receive much less attention. In order to tackle these challenges, a privacy‐preserving and device‐managed task offloading scheme is proposed in this paper for MEC. This scheme can achieve near‐optimal latency and energy performance while protecting the location privacy and usage pattern privacy of users. Firstly, we formulate the joint optimization problem of task offloading and privacy preservation as a semiparametric contextual multi‐armed bandit (MAB) problem, which has a relaxed reward model. Then, we propose a privacy‐aware online task offloading (PAOTO) algorithm based on the transformed Thompson sampling (TS) architecture, through which we can (1) receive the best possible delay and energy consumption performance, (2) achieve the goal of preserving privacy, and (3) obtain an online device‐managed task offloading policy without requiring any system‐level information. Simulation results demonstrate that the proposed scheme outperforms the existing methods in terms of minimizing the system cost and preserving the privacy of users.

Queueing analysis of GPU-based inference servers with dynamic batching: A closed-form characterization

GPU-accelerated computing is a key technology to realize high-speed inference servers using deep neural networks (DNNs). An important characteristic of GPU-based inference is that the computational efficiency, in terms of the processing speed and energy consumption, drastically increases by processing multiple jobs together in a batch. In this paper, we formulate GPU-based inference servers as a batch service queueing model with batch-size dependent processing times. We first show that the energy efficiency of the server monotonically increases with the arrival rate of inference jobs, which suggests that it is energy-efficient to operate the inference server under a utilization level as high as possible within a latency requirement of inference jobs. We then derive a closed-form upper bound for the mean latency, which provides a simple characterization of the latency performance. Through simulation and numerical experiments, we show that the exact value of the mean latency is well approximated by this upper bound. We further compare this upper bound with the latency curve measured in real implementation of GPU-based inference servers and we show that the real performance curve is well explained by the derived simple formula.

MDev-NVMe: Mediated Pass-Through NVMe Virtualization Solution With Adaptive Polling

The fast access to data and high parallel processing in high-performance computing instigates an urgent demand on the improvement of the NVMe storage within modern data centers. However, the former NVMe virtualization&#x2019;s unsatisfactory performance demonstrates that NVMe devices are often underutilized within cloud platforms. An NVMe virtualization mechanism with high performance and device sharing has captured researchers and developers&#x2019; attention. This article introduces MDev-NVMe, a new virtualization solution for NVMe storage device with (1) full NVMe storage virtualization for VMs running native NVMe driver, (2) a mediated pass-through mechanism for NVMe management, and (3) adaptive configuration of active polling optimization to simultaneously achieve high throughput, low latency performance, and substantial device scalability. We practically implement the MDev-NVMe as a Linux kernel module. This article subsequently evaluates MDev-NVMe with Intel OPTANE and P3600 SSD by comparing several mainstream NVMe virtualization mechanisms using application-level I/O benchmarks. MDev-NVMe with active polling can demonstrate a 142 percent improvement over native (interrupt-driven) throughput and over 2.5 &#x00D7; the <i>Virtio</i> throughput with only 70 percent native average latency and 31 percent <i>Virtio</i> average latency. Finally, the advantages of MDev-NVMe and the importance of adaptive polling are discussed, offering evidence that MDev-NVMe is a superior virtualization choice for cloud storage.

The Blind Side: Latency Challenges in Millimeter Wave Networks for Connected Vehicle Applications

Millimeter-wave (mmWave) is a promising network access technology to enable the high data rates, high reliability and ultra-low latency required by connected vehicle services in future vehicular networks. However, mmWave links are prone to blockages due to high penetration loss, which can cause frequent service interruptions and degrade the system performance in terms of reliability and latency. In this study, we analyze the latency and reliability performance of mmWave communications between vehicles and roadside units (RSUs) in a highway scenario, where the line-of-sight (LOS) links can be blocked by vehicles. First, we establish a continuous-time Markov chain model of the blockage events. By using the steady-state solution of this model, we explicitly derive the blockage probability and average blockage duration, which can be used to characterize the reliability and latency performance. We validate the accuracy of the analytical model by comparing it with simulations using real-world traffic data and vehicle distributions. We demonstrate that reducing the duration of long-lasting blockage events is more challenging than reducing the frequency of blockages. We consider three approaches to control the blockage probability and distribution of blockage durations: (i) increasing RSU density, (ii) increasing RSU heights and (iii) managing the vehicular speed limits. We show that the first two approaches are effective in reducing the blockage probability, whereas the third approach can be used to eliminate long blockages and improve latency performance. We discuss the implications of our results in terms of the benefits and challenges associated with these approaches.

Distributed Multi-Cloud Multi-Access Edge Computing by Multi-Agent Reinforcement Learning

In this paper, we consider a three-layer distributed multi-access edge computing (MEC) network where multiple clouds, MEC servers, and edge devices (EDs) are deployed at the top layer, middle layer, and bottom layer, respectively. Each cloud center (CC) is associated with an independent service provider and publishes an application-driven computing task. To deliver the tasks, CCs rely on EDs to generate the raw data and offload part of the computing tasks to both EDs and MEC servers such that their computing and transmission resources can be fully utilized to reduce the system latency. However, in such a three-layer network, the distributed deployment of tasks leads to inevitable resource competition among CCs. To address this issue, we propose a distributed scheme based on multi-agent reinforcement learning, where each CC jointly determines the task offloading and resource allocation strategy based on its inference of other CCs' decisions. Simulation results indicate that a lower system latency is achieved via our proposed scheme compared with the existing schemes. In addition, the influence of the number of CCs, MEC servers, and EDs on latency performance is also discussed.

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.

Real-Time Demand Response Using NB-IoT

The Internet of Things (IoT) already connects billions of devices and keeps growing exponentially. These devices are designed to be integrated with industrial machines, home appliances, and infrastructures. One such use case is to build a smart grid management system that relies on demand-response techniques to control the appliances automatically so that the power can be distributed optimally. The mission-critical smart grid communications require secure, reliable, two-way communicable, and latency bounded connections between the management system and the electrical appliances. To realize these, cellular technology is arguably the most feasible solution. 3GPP has already released the narrowband-IoT (NB-IoT) standards as the low-power dense-area coverage IoT cellular solution. In this article, we present an NB-IoT system to monitor and control the connected electrical appliances in a smart grid network. The platform is also capable of configuring the network dynamically. We assess the latency performance for our solution using the commercially available Orange network in Belgium. It is observed that NB-IoT enabled devices can be controlled and monitored with a maximum latency less than 8 s in the deep-indoor environment and within 2 s for the outdoor environment.

Latency performance modeling and analysis for hyperledger fabric blockchain network

Blockchain has been one of the most attractive technologies for many modern and even future applications. Fabric, an open-source framework to implement the permissioned enterprise-grade blockchain, is getting increasing attention from innovators. The latency performance is crucial to the Fabric blockchain in assessing its effectiveness. Many empirical studies were conducted to analyze this performance based on different hardware platforms. These experimental results are not comparable as they are highly dependent on the underlying networks. Moreover, theoretical analysis on the latency of Fabric blockchain still receives much less attention. This paper provides a novel theoretical model to calculate the transaction latency under various network configurations such as block size, block interval, etc. Subsequently, we validate the proposed latency model with experiments, and the results show that the difference between analytical and experimental results is as low as 6.1%. We also identify some performance bottlenecks and give insights from the developer’s perspective.

A Photonic Recurrent Neuron for Time-Series Classification

Neuromorphic photonics has turned into a key research area for enabling neuromorphic computing at much higher data-rates compared to their electronic counterparts, improving significantly the (multiply-and-accumulate) MAC/sec. At the same time, time-series classification problems comprise a large class of artificial intelligence (AI) applications where speed and latency can have a decisive role in their hardware deployment roadmap, highlighting the need for ultra-fast hardware implementations of simplified recurrent neural networks (RNN) that can be extended in more advanced long-short-term-memory (LSTM) and gated recurrent unit (GRU) machines. Herein, we experimentally demonstrate a novel photonic recurrent neuron (PRN) to classify successfully a time-series vector with 100-psec optical pulses and up to 10 Gb/s data speeds, reporting on the fastest all-optical real-time classifier. Experimental classification of 3-bit optical binary data streams is presented, revealing an average accuracy of >91% and confirming the potential of PRNs to boost speed and latency performance in time-series AI applications.

Comparative Analysis of Spintronic Memories for Low Power on-chip Caches

The continuous downscaling in CMOS devices has increased leakage power and limited the performance to a few GHz. The research goal has diverted from operating at high frequencies to deliver higher performance in essence with lower power. CMOS based on-chip memories consumes significant fraction of power in modern processors. This paper aims to explore the suitability of beyond CMOS, emerging magnetic memories for the use in memory hierarchy, attributing to their remarkable features like nonvolatility, high-density, ultra-low leakage and scalability. NVSim, a circuit-level tool, is used to explore different design layouts and memory organizations and then estimate the energy, area and latency performance numbers. A detailed system-level performance analysis of STT-MRAM and SOT-MRAM technologies and comparison with 22[Formula: see text]nm SRAM technology are presented. Analysis infers that in comparison to the existing 22[Formula: see text]nm SRAM technology, SOT-MRAM is more efficient in area for memory size [Formula: see text][Formula: see text]KB, speed and energy consumption for cache size [Formula: see text][Formula: see text]KB. A typical 256[Formula: see text]KB SOT-MRAM cache design is 27.74% area efficient, 2.97 times faster and consumes 76.05% lesser leakage than SRAM counterpart and these numbers improve for larger cache sizes. The article deduces that SOT-MRAM technology has a promising potential to replace SRAM in lower levels of computer memory hierarchy.

Performance Evaluation of an Effective Mobility Model for D2D Communications

Device-to-Device (D2D) communication, in which devices take advantage of their physical proximity, becomes an emerging technology and a hot research area for 5G cellular networks. Many research works have thoroughly tried to deal with interference management, resource allocation, security, opportunistic device contacts and user mobility. However, unpredictable device mobility may influence the service continuity of a D2D communication. Therefore, the effects of the mobility model on D2D communication performances must be fully studied. In this paper, we adapt the RPGM (Reference Point Group Mobility), a group mobility model, to the D2D communication. It is suitable for modelling the D2D mobility and manages the relationship among mobile devices. To demonstrate the effectiveness of our adopted mobility model, we compare it to other popular mobility models. As expected, our adopted mobility model outperforms the other models in terms of handover latency and system performances with different D2D applications.

Achieving Low-Latency Human-to-Machine (H2M) Applications: An Understanding of H2M Traffic for AI-Facilitated Bandwidth Allocation

Human-controlled and haptic feedback data in emerging Tactile Internet human-to-machine (H2M) applications require stringent low-latency transmission. Understanding the traffic features of the new applications is vital in innovating network control and resource allocation strategies to meet their latency demand. In this article, we present our experimental study on human control and haptic feedback traffic in H2M applications and investigate novel bandwidth allocation schemes in supporting converged H2M application delivery over access networks. We introduce our haptic experiment system, the developed H2M applications, and analyze the control and feedback traffic traces collected. Then, exploiting the correlation between real-time control and feedback reported in our analysis, we propose an artificial intelligence-facilitated low-latency bandwidth allocation (ALL) scheme for emerging H2M applications. ALL provisions priority-differentiated bandwidth allocation for aggregated H2M and conventional content-centric applications over future access networks. By using ALL, the central office preallocates bandwidth for control and its corresponding feedback traffic interactively and prioritizes their transmission over content traffic. This expedites H2M application delivery by eliminating the report-then-grant process in the existing bandwidth allocation schemes. Via extensive simulations injected with experimental traffic traces, we comprehensively investigate the latency performance of ALL and existing schemes. Our results validate the superior capability of ALL in reducing and constraining latency for H2M applications.