Meet Latency Requirements Research Articles

Federated In-Network Machine Learning for Privacy-Preserving IoT Traffic Analysis

The expanding use of Internet-of-Things (IoT) has driven machine learning (ML)-based traffic analysis. 5G networks’ standards, requiring low-latency communications for time-critical services, pose new challenges to traffic analysis. They necessitate fast analysis and response, preventing service disruption or security impact on network infrastructure. Distributed intelligence on IoT edge has been studied to analyze traffic, but introduces delays and raises privacy concerns. Federated learning can address privacy concerns, but does not meet latency requirements. In this article, we propose FLIP4: an efficient federated learning-based framework for in-network traffic analysis. Our solution introduces a lightweight federated tree-based model, offloaded and running within network devices. FLIP4 consumes less resources than previous solutions and reduces communication overheads, making it well-suited for IoT edge traffic analysis. It ensures prompt mitigation and minimal impact on services in the presence of false alerts using two approaches (metering and dropping), thereby balancing learning accuracy and privacy requirements.

LESS-ON: Load-aware edge server shutdown for energy saving in cellular networks

While advances in wireless networks enable novel services with previously unreachable latency guarantees, edge computing becomes essential for delivering computing resources close to the users and meeting the strict latency requirements. However, addressing the energy footprint of computing resources is crucial amid the pressing sustainability concerns. The energy consumption of idle resources accounts for a significant part of the total energy footprint. While server shutdown during low-demand periods is common in cloud computing, it is challenging to determine which edge servers to shut down and how to route requests due to the stringent latency requirements of the applications. Thus, this work formulates an optimal orchestration policy to minimize the energy consumption of the edge computing infrastructure and presents LESS-ON, a strategy with a polynomial time complexity that reduces the operational energy footprint of edge computing by shutting down edge servers during low-demand periods. In contrast to previous studies, LESS-ON considers the energy requirements associated with routing requests to the designated edge servers. Our numerical evaluation shows that LESS-ON reduces the total consumption by 42% with respect to the common always-on approach during low-demand periods and by 35% over 24 h, all while meeting latency requirements.

Data Aggregation through Hybrid Optimal Probability in Wireless Sensor Networks


 INTRODUCTION: In the realm of Wireless Sensor Networks (WSN), effective data dissemination is vital for applications like traffic alerts, necessitating innovative solutions to tackle challenges such as broadcast storms.
 OBJECTIVES: This paper proposes a pioneering framework that leverages probabilistic data aggregation to optimize communication efficiency and minimize redundancy.
 METHODS: The proposed adaptable system extracts valuable insights from the knowledge base, enabling dynamic route adjustments based on application-specific criteria. Through simulations addressing bandwidth limitations and local broadcast issues, we establish a robust WSN-based traffic information system.
 RESULTS: By employing primal-dual decomposition, the proposed approach identifies optimal packet aggregation probabilities and durations, resulting in reduced energy consumption while meeting latency requirements.
 CONCLUSION: The efficacy of proposed method is demonstrated across various traffic and topology scenarios, affirming that probabilistic data aggregation effectively mitigates the local broadcast problem, ultimately leading to decreased bandwidth demands.

Evaluating virtualization for fog monitoring of real-time applications in mixed-criticality systems

Technological advances in embedded systems and the advent of fog computing led to improved quality of service of applications of cyber-physical systems. In fact, the deployment of such applications on powerful and heterogeneous embedded systems, such as multiprocessors system-on-chips (MPSoCs), allows them to meet latency requirements and real-time operation. Highly relevant to the industry and our reference case-study, the challenging field of nuclear fusion deploys the aforementioned applications, involving high-frequency control with hard real-time and safety constraints. The use of fog computing and MPSoCs is promising to achieve safety, low latency, and timeliness of such control. Indeed, on one hand, applications designed according to fog computing distribute computation across hierarchically organized and geographically distributed edge devices, enabling timely anomaly detection during high-frequency sampling of time series, and, on the other hand, MPSoCs allow leveraging fog computing and integrating monitoring by deploying tasks on a flexible platform suited for mixed-criticality software, leading to so-called mixed criticality systems (MCSs). However, the integration of such software on the same MPSoC opens challenges related to predictability and reliability guarantees, as tasks interfering with each other when accessing the same shared MPSoC resources may introduce non-deterministic latency, possibly leading to failures on account of deadline overruns. Addressing the design, deployment, and evaluation of MCSs on MPSoCs, we propose a model-based system development process that facilitates the integration of real-time and monitoring software on the same platform by means of a formal notation for modeling the design and deployment of MPSoCs. The proposed notation allows developers to leverage embedded hypervisors for monitoring real-time applications and guaranteeing predictability by isolation of hardware resources. Providing evidence of the feasibility of our system development process and evaluating the industry-relevant class of nuclear fusion applications, we experiment with a safety-critical case-study in the context of the ITER nuclear fusion reactor. Our experimentation involves the design and evaluation of several prototypes deployed as MCSs on a virtualized MPSoC, showing that deployment choices linked to the monitor placement and virtualization configurations (e.g., resource allocation, partitioning, and scheduling policies) can significantly impact the predictability of MCSs in terms of Worst-Case Execution Times and other related metrics.

FogWise: On the limits of the coexistence of heterogeneous applications on Fog computing and Internet of Vehicles

AbstractMobile devices, Internet of Things applications, and connected vehicles in the Internet of Vehicles (IoV) are proliferating. Recently, vehicles have emerged as a possibility to become communication and processing infrastructures called vehicular Fog computing (VFC). Thus, science and industry forecast increasing demand for communication and cloud infrastructures to deal with data traffic, processing, and storage. Still, they do not evaluate that properly. This work investigates how heterogeneity affects the ability of IoV and VFC to meet application requirements. We leverage Fog computing, which extends Edge and Cloud computing and complements them, especially to meet str applications with strict latency requirements. We built a simulation model that evaluates how vehicles behave in such an environment. We argue that vehicles need to follow request‐processing‐response‐actuation programming model based on distributed auction protocol to match clients and servers and still meet latency requirements. We extend and propose requirements that both applications and architectures should meet. Preliminary results indicate thresholds in terms of vehicle density, request rate, request size, client‐server ratio, and auction method in which VFC can operate to meet application requirements under mobility.

A Green Converged TWDM-PON and 5G HetNet Catering Applications Demanding Low Latency

Abstract Integration of Passive Optical Network (PON) and wireless network technologies can offer a high data rate with ubiquitous Internet access to the end-users. Due to the growing importance of reducing the carbon footprint of telecommunication networks, energy-saving mechanisms are dispensed in the networking equipment. One of the commonly used approaches to reduce energy consumption is the “sleep mode” in which a network equipment periodically turns off some of its components so as to reduce energy consumption. Note that, the current PON and wireless access network technologies have their own power-saving modes. In order to gain power-saving and meet latency of applications in an integrated PON and wireless access network technologies, both optical and wireless access technologies need to have consent on how traffic management (bandwidth allocation, traffic forwarding path), as well as power-saving modes, should be managed in the network. That is, there needs to have synchronization of these two network segments. Otherwise, the end-users would experience longer delay and lower throughput. In this paper, we introduce a Delay-Aware Integrated Sleep Mode (DISM) solution designed considering 5G Heterogeneous Networks (HetNets) architecture which connected with core networks through Time and Wavelength Division Multiplexed Passive Optical Network (TWDM-PON). In particular, we propose a power saving mechanism using sleep mode for TWDM-PON taking into account packet forwarding delay from the optical network termination point to the end-users’ mobile terminal so as to maximize energy-saving in TWDM-PON while meeting latency requirements of applications with stringent latency requirements. Results obtained through simulations impart that our proposed scheme can significantly satisfy the delay requirement of the traffic while it can still reduce energy consumption in TWDM-PON. Furthermore, the results provide insights on how power saving mechanism in PON should be managed for such 5G HetNet and optical integrated networks in order to satisfy the latency requirement of the traffic.

Collaborative resource allocation for Cloud of Things systems

The conceptual approach known as Fog/Edge Computing has recently emerged, aiming to move part of the computing and storage resources from the cloud to the edge of the network. The combination of IoT devices, edge nodes, and the Cloud gives rise to a three-tier Cloud of Things (CoT) architecture. In the complex and dynamic CoT ecosystems, a key issue is how to efficiently and effectively allocate resources to meet the demands of applications. Similar to traditional clouds, the goal of resource allocation in the CoT is to maximize the number of applications served by the infrastructure while ensuring a target operational cost. We propose a resource allocation algorithm for CoT systems that (i) supports heterogeneity of devices and applications, (ii) leverages the distributed nature of edge nodes to promote collaboration during the allocation process and (iii) provides an efficient usage of the system resources while meeting latency requirements and considering different priorities of IoT applications. Our algorithm follows a heuristic-based approach inspired on an economic model for solving the resource allocation problem in CoT. A set of simulations were performed, with promising results, showing that our collaborative resource allocation algorithm is more scalable, reduces the response time for applications and the energy consumption of end devices, in comparison to a two-tier, Cloud-based approach. Moreover, the network traffic between edge nodes, and between the Edge and Cloud tiers, is considerably smaller when using our collaborative solution, in comparison to other evaluated approaches.

Enabling hard service guarantees in Software-Defined Smart Grid infrastructures

Information and Communication Technology (ICT) infrastructures play a key role in the evolution from traditional power systems to Smart Grids. Increasingly fluctuating power flows, sparked by the transition towards sustainable energy generation, become a major issue for power grid stability. To deal with this challenge, future Smart Grids require precise monitoring and control, which in turn demand for reliable, real-time capable and cost-efficient communications. For this purpose, we propose applying Software-Defined Networking (SDN) to handle the manifold requirements of Smart Grid communications. To achieve reliability, our approach encompasses fast recovery after failures in the communication network and dynamic service-aware network (re-)configuration. Network Calculus (NC) logic is embedded into our SDN controller for meeting latency requirements imposed by the standard IEC 61850 of the International Electrotechnical Commission (IEC). Thus, routing provides delay-optimal paths under consideration of existing cross traffic. Also, continuous latency bound compliance is ensured by combining NC delay supervision with means of flexible reconfiguration. For evaluation we consider the well-known Nordic 32 test system, on which we map a corresponding communication network in both experiment and emulation. The described functionalities are validated, employing realistic IEC 61850 transmissions and distributed control traffic. Our results show that hard service guarantees can be ensured with the help of the proposed SDN solution. On this basis, we derive extremely time critical services, which must not be subjected to flexible reconfiguration.

CALOREE

Many modern computing systems must provide reliable latency with minimal energy. Two central challenges arise when allocating system resources to meet these conflicting goals: (1) complexity modern hardware exposes diverse resources with complicated interactions and (2) dynamics latency must be maintained despite unpredictable changes in operating environment or input. Machine learning accurately models the latency of complex, interacting resources, but does not address system dynamics; control theory adjusts to dynamic changes, but struggles with complex resource interaction. We therefore propose CALOREE, a resource manager that learns key control parameters to meet latency requirements with minimal energy in complex, dynamic en- vironments. CALOREE breaks resource allocation into two sub-tasks: learning how interacting resources affect speedup, and controlling speedup to meet latency requirements with minimal energy. CALOREE deines a general control system whose parameters are customized by a learning framework while maintaining control-theoretic formal guarantees that the latency goal will be met. We test CALOREE's ability to deliver reliable latency on heterogeneous ARM big.LITTLE architectures in both single and multi-application scenarios. Compared to the best prior learning and control solutions, CALOREE reduces deadline misses by 60% and energy consumption by 13%.

Integrated 3D-stacked server designs for increasing physical density of key-value stores

Key-value stores, such as Memcached, have been used to scale web services since the beginning of the Web 2.0 era. Data center real estate is expensive, and several industry experts we have spoken to have suggested that a significant portion of their data center space is devoted to key value stores. Despite its wide-spread use, there is little in the way of hardware specialization for increasing the efficiency and density of Memcached; it is currently deployed on commodity servers that contain high-end CPUs designed to extract as much instruction-level parallelism as possible. Out-of-order CPUs, however have been shown to be inefficient when running Memcached. To address Memcached efficiency issues, we propose two architectures using 3D stacking to increase data storage efficiency. Our first 3D architecture, Mercury, consists of stacks of ARM Cortex-A7 cores with 4GB of DRAM, as well as NICs. Our second architecture, Iridium, replaces DRAM with NAND Flash to improve density. We explore, through simulation, the potential efficiency benefits of running Memcached on servers that use 3D-stacking to closely integrate low-power CPUs with NICs and memory. With Mercury we demonstrate that density may be improved by 2.9X, power efficiency by 4.9X, throughput by 10X, and throughput per GB by 3.5X over a state-of-the-art server running optimized Memcached. With Iridium we show that density may be increased by 14X, power efficiency by 2.4X, and throughput by 5.2X, while still meeting latency requirements for a majority of requests.

Integrated 3D-stacked server designs for increasing physical density of key-value stores

Key-value stores, such as Memcached, have been used to scale web services since the beginning of the Web 2.0 era. Data center real estate is expensive, and several industry experts we have spoken to have suggested that a significant portion of their data center space is devoted to key value stores. Despite its wide-spread use, there is little in the way of hardware specialization for increasing the efficiency and density of Memcached; it is currently deployed on commodity servers that contain high-end CPUs designed to extract as much instruction-level parallelism as possible. Out-of-order CPUs, however have been shown to be inefficient when running Memcached. To address Memcached efficiency issues, we propose two architectures using 3D stacking to increase data storage efficiency. Our first 3D architecture, Mercury, consists of stacks of ARM Cortex-A7 cores with 4GB of DRAM, as well as NICs. Our second architecture, Iridium, replaces DRAM with NAND Flash to improve density. We explore, through simulation, the potential efficiency benefits of running Memcached on servers that use 3D-stacking to closely integrate low-power CPUs with NICs and memory. With Mercury we demonstrate that density may be improved by 2.9X, power efficiency by 4.9X, throughput by 10X, and throughput per GB by 3.5X over a state-of-the-art server running optimized Memcached. With Iridium we show that density may be increased by 14X, power efficiency by 2.4X, and throughput by 5.2X, while still meeting latency requirements for a majority of requests.

Recovering From a Node Failure in Wireless Sensor-Actor Networks With Minimal Topology Changes

In wireless sensor-actor networks, sensors probe their surroundings and forward their data to actor nodes. Actors collaboratively respond to achieve predefined application mission. Since actors have to coordinate their operation, it is necessary to maintain a strongly connected network topology at all times. Moreover, the length of the inter-actor communication paths may be constrained to meet latency requirements. However, a failure of an actor may cause the network to partition into disjoint blocks and would, thus, violate such a connectivity goal. One of the effective recovery methodologies is to autonomously reposition a subset of the actor nodes to restore connectivity. Contemporary recovery schemes either impose high node relocation overhead or extend some of the inter-actor data paths. This paper overcomes these shortcomings and presents a Least-Disruptive topology Repair (LeDiR) algorithm. LeDiR relies on the local view of a node about the network to devise a recovery plan that relocates the least number of nodes and ensures that no path between any pair of nodes is extended. LeDiR is a localized and distributed algorithm that leverages existing route discovery activities in the network and imposes no additional prefailure communication overhead. The performance of LeDiR is analyzed mathematically and validated via extensive simulation experiments.

Optimizing latency and throughput of application workflows on clusters

Scheduling, in many application domains, involves optimization of multiple performance metrics. For example, application workflows with real-time constraints have strict throughput requirements and also desire a low latency or response time. In this paper, we present a novel algorithm for the scheduling of workflows that act on a stream of input data. Our algorithm focuses on the two performance metrics, latency and throughput, and minimizes the latency of workflows while satisfying strict throughput requirements. We also describe steps to use the above approach to solve the problem of meeting latency requirements while maximizing throughput. We leverage pipelined, task and data parallelism in a coordinated manner to meet these objectives and investigate the benefit of task duplication in alleviating communication overheads in the pipelined schedule for different workflow characteristics. The proposed algorithm is designed for a realistic bounded multi-port communication model, where each processor can simultaneously communicate with at most k distinct processors. Experimental evaluation using synthetic benchmarks as well as those derived from real applications shows that our algorithm consistently produces lower latency schedules that meet throughput requirements, even when previously proposed schemes fail.