On the Analysis of Inter-Relationship between Auto-Scaling Policy and QoS of FaaS Workloads.

A recent development in cloud computing has introduced serverless technology, enabling the convenient and flexible management of cloud-native applications. Typically, the Function-as-a-Service (FaaS) solutions rely on serverless backend solutions, such as Kubernetes (K8s) and Knative, to leverage the advantages of resource management for underlying containerized contexts, including auto-scaling and pod scheduling. To take the advantages, recent cloud service providers also deploy self-hosted serverless services by facilitating their on-premise hosted FaaS platforms rather than relying on commercial public cloud offerings. However, the lack of standardized guidelines on K8s abstraction to fairly schedule and allocate resources on auto-scaling configuration options for such on-premise hosting environment in serverless computing poses challenges in meeting the service level objectives (SLOs) of diverse workloads. This study fills this gap by exploring the relationship between auto-scaling behavior and the performance of FaaS workloads depending on scaling-related configurations in K8s. Based on comprehensive measurement studies, we derived the logic as to which workload should be applied and with what type of scaling configurations, such as base metric, threshold to maximize the difference in latency SLO, and number of responses. Additionally, we propose a methodology to assess the scaling efficiency of the related K8s configurations regarding the quality of service (QoS) of FaaS workloads.

ScalableDigitalHealth (SDH): An IoT-Based Scalable Framework for Remote Patient Monitoring.

Addressing the increasing demand for remote patient monitoring, especially among the elderly and mobility-impaired, this study proposes the "ScalableDigitalHealth" (SDH) framework. The framework integrates smart digital health solutions with latency-aware edge computing autoscaling, providing a novel approach to remote patient monitoring. By leveraging IoT technology and application autoscaling, the "SDH" enables the real-time tracking of critical health parameters, such as ECG, body temperature, blood pressure, and oxygen saturation. These vital metrics are efficiently transmitted in real time to AWS cloud storage through a layered networking architecture. The contributions are two-fold: (1) establishing real-time remote patient monitoring and (2) developing a scalable architecture that features latency-aware horizontal pod autoscaling for containerized healthcare applications. The architecture incorporates a scalable IoT-based architecture and an innovative microservice autoscaling strategy in edge computing, driven by dynamic latency thresholds and enhanced by the integration of custom metrics. This work ensures heightened accessibility, cost-efficiency, and rapid responsiveness to patient needs, marking a significant leap forward in the field. By dynamically adjusting pod numbers based on latency, the system optimizes system responsiveness, particularly in edge computing's proximity-based processing. This innovative fusion of technologies not only revolutionizes remote healthcare delivery but also enhances Kubernetes performance, preventing unresponsiveness during high usage.

An Autoscaling System Based on Predicting the Demand for Resources and Responding to Failure in Forecasting.

In recent years, the convergence of edge computing and sensor technologies has become a pivotal frontier revolutionizing real-time data processing. In particular, the practice of data acquisition-which encompasses the collection of sensory information in the form of images and videos, followed by their transmission to a remote cloud infrastructure for subsequent analysis-has witnessed a notable surge in adoption. However, to ensure seamless real-time processing irrespective of the data volume being conveyed or the frequency of incoming requests, it is vital to proactively locate resources within the cloud infrastructure specifically tailored to data-processing tasks. Many studies have focused on the proactive prediction of resource demands through the use of deep learning algorithms, generating considerable interest in real-time data processing. Nonetheless, an inherent risk arises when relying solely on predictive resource allocation, as it can heighten the susceptibility to system failure. In this study, a framework that includes algorithms that periodically monitor resource requirements and dynamically adjust resource provisioning to match the actual demand is proposed. Under experimental conditions with the Bitbrains dataset, setting the network throughput to 300 kB/s and with a threshold of 80%, the proposed system provides a 99% performance improvement in terms of the autoscaling algorithm and requires only 0.43 ms of additional computational overhead compared to relying on a simple prediction model alone.

Performance-Cost Trade-Off in Auto-Scaling Mechanisms for Cloud Computing.

Cloud computing has been widely adopted over the years by practitioners and companies with a variety of requirements. With a strong economic appeal, cloud computing makes possible the idea of computing as a utility, in which computing resources can be consumed and paid for with the same convenience as electricity. One of the main characteristics of cloud as a service is elasticity supported by auto-scaling capabilities. The auto-scaling cloud mechanism allows adjusting resources to meet multiple demands dynamically. The elasticity service is best represented in critical web trading and transaction systems that must satisfy a certain service level agreement (SLA), such as maximum response time limits for different types of inbound requests. Nevertheless, existing cloud infrastructures maintained by different cloud enterprises often offer different cloud service costs for equivalent SLAs upon several factors. The factors might be contract types, VM types, auto-scaling configuration parameters, and incoming workload demand. Identifying a combination of parameters that results in SLA compliance directly in the system is often sophisticated, while the manual analysis is prone to errors due to the huge number of possibilities. This paper proposes the modeling of auto-scaling mechanisms in a typical cloud infrastructure using a stochastic Petri net (SPN) and the employment of a well-established adaptive search metaheuristic (GRASP) to discover critical trade-offs between performance and cost in cloud services.The proposed SPN models enable cloud designers to estimate the metrics of cloud services in accordance with each required SLA such as the best configuration, cost, system response time, and throughput.The auto-scaling SPN model was extensively validated with 95% confidence against a real test-bed scenario with 18.000 samples. A case-study of cloud services was used to investigate the viability of this method and to evaluate the adoptability of the proposed auto-scaling model in practice. On the other hand, the proposed optimization algorithm enables the identification of economic system configuration and parameterization to satisfy required SLA and budget constraints. The adoption of the metaheuristic GRASP approach and the modeling of auto-scaling mechanisms in this work can help search for the optimized-quality solution and operational management for cloud services in practice.

Framework for Efficient Auto-Scaling of Virtual Network Functions in a Cloud Environment.

Network Function Virtualization (NFV) offers an alternate method to design, deploy and manage network services. The NFV decouples network functions from the dedicated hardware and moves them to the virtual servers so that they can run in the software. One of the major strengths of the NFV is its ability to dynamically extend or reduce resources allocated to Virtual Network Functions (VNF) as needed and at run-time. There is a need for a comprehensive metering component in the cloud to store and process the metrics/samples for efficient auto-scaling or load-management of the VNF. In this paper, we propose an integrating framework for efficient auto-scaling of VNF using Gnocchi; a time-series database that is integrated within the framework to store, handle and index the time-series data. The objective of this study is to validate the efficacy of employing Gnocchi for auto-scaling of VNF, in terms of aggregated data points, database size, data recovery speed, and memory consumption. The employed methodology is to perform a detailed empirical analysis of the proposed framework by deploying a fully functional cloud to implement NFV architecture using several OpenStack components including Gnocchi. Our results show a significant improvement over the legacy Ceilometer configuration in terms of lower metering storage size, less memory utilization in processing and management of metrics, and reduced time delay in retrieving the monitoring data to evaluate alarms for the auto-scaling of VNF.

An SDN-Based Solution for Horizontal Auto-Scaling and Load Balancing of Transparent VNF Clusters.

This paper studies the problem of the dynamic scaling and load balancing of transparent virtualized network functions (VNFs). It analyzes different particularities of this problem, such as loop avoidance when performing scaling-out actions, and bidirectional flow affinity. To address this problem, a software-defined networking (SDN)-based solution is implemented consisting of two SDN controllers and two OpenFlow switches (OFSs). In this approach, the SDN controllers run the solution logic (i.e., monitoring, scaling, and load-balancing modules). According to the SDN controllers instructions, the OFSs are responsible for redirecting traffic to and from the VNF clusters (i.e., load-balancing strategy). Several experiments were conducted to validate the feasibility of this proposed solution on a real testbed. Through connectivity tests, not only could end-to-end (E2E) traffic be successfully achieved through the VNF cluster, but the bidirectional flow affinity strategy was also found to perform well because it could simultaneously create flow rules in both switches. Moreover, the selected CPU-based load-balancing method guaranteed an average imbalance below 10% while ensuring that new incoming traffic was redirected to the least loaded instance without requiring packet modification. Additionally, the designed monitoring function was able to detect failures in the set of active members in near real-time and active new instances in less than a minute. Likewise, the proposed auto-scaling module had a quick response to traffic changes. Our solution showed that the use of SDN controllers along with OFS provides great flexibility to implement different load-balancing, scaling, and monitoring strategies.

Horizontal Pod Autoscaling in Kubernetes for Elastic Container Orchestration.

Kubernetes, an open-source container orchestration platform, enables high availability and scalability through diverse autoscaling mechanisms such as Horizontal Pod Autoscaler (HPA), Vertical Pod Autoscaler and Cluster Autoscaler. Amongst them, HPA helps provide seamless service by dynamically scaling up and down the number of resource units, called pods, without having to restart the whole system. Kubernetes monitors default Resource Metrics including CPU and memory usage of host machines and their pods. On the other hand, Custom Metrics, provided by external software such as Prometheus, are customizable to monitor a wide collection of metrics. In this paper, we investigate HPA through diverse experiments to provide critical knowledge on its operational behaviors. We also discuss the essential difference between Kubernetes Resource Metrics (KRM) and Prometheus Custom Metrics (PCM) and how they affect HPA's performance. Lastly, we provide deeper insights and lessons on how to optimize the performance of HPA for researchers, developers, and system administrators working with Kubernetes in the future.

An M/PH/1 queue with workload-dependent processing speed and vacations.

Motivated by the trade-off issue between delay performance and energy consumption in modern computer and communication systems, we consider a single-server queue with phase-type service requirements and with the following two special features: Firstly, the service speed is a piecewise constant function of the workload. Secondly, the server switches off when the system becomes empty, only to be activated again when the workload reaches a certain threshold. For this system, we obtain the steady-state workload distribution and its moments of any order. We use this result to choose the activation threshold such that a certain cost function, involving processing costs, activation costs and mean workload, is minimized.

The time-dependent synergism of the six-component mixtures of substituted phenols, pesticides and ionic liquids to Caenorhabditis elegans.

Traditional environmental risk assessment rarely focused on exposures to multi-component mixtures which may cause toxicological interactions and usually ignored that toxicity is a process in time, which may underestimate the environment risk of mixtures. In this paper, six chemicals belonging to three categories, two substituted phenols, two pesticides and two Ionic liquids, were picked to construct a six-component mixture system. To systematically examine the effects of various concentration compositions, the uniform design ray method was employed to design nine mixture rays with nine mixture ratios and for every mixture ray 12 concentration levels were specified by the fixed ratio ray design. The improved combination index was used to evaluate the combined toxicities of the mixtures to Caenorhabditis elegans (C. elegans) in the exposure times of 6, 12 and 24h. It was shown that the mixture rays display time-dependent synergism, i.e. the range of synergistic effect narrows and the strength of synergism runs down with exposure time, which illustrates that the mixture toxicity of some chemicals is not a sum of individual toxicities at some exposure times and it is necessary to consider the toxicological interaction in mixtures.