Predictive Autoscaling of Microservices Hosted in Fog Microdata Center

Abdullah, Muhammad; Iqbal, Waheed; Mahmood, Arif; Bukhari, Faisal; Erradi, Abdelkarim

doi:10.1109/jsyst.2020.2997518

Cited by 40 publications

(18 citation statements)

References 44 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This experiment evaluated the proposed distributed model learning effectiveness and compared it with a single standalone learning model ( 1). We focused on two different scenarios, where the processes can work in multiple CPUs in the same place (e.g., Tensorflow working with multiple CPUs in the same machine), or a scenario where CPUs are disaggregated and become independent of each other (e.g., different Edge nodes cooperating).…”

Section: B Comparison Of Single and Distributed Modelsmentioning

confidence: 99%

“…In contrast, emerging scenarios like the IoT, smart cities, domotic, intelligent surveillance, and e-healthcare usually require proximity and quick reaction time while generating massive amounts of data transmitted to the analytics applications. Fog computing is more attractive for such demand [1]. Fog computing takes the computation to the Edge, moving data processing close to the sources, and reducing data to synthesized volumes to be transmitted north-bound to the Cloud, as shown in Figure 1.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Automatic Distributed Deep Learning Using Resource-Constrained Edge Devices

Torre

Bahadori

Baig

et al. 2022

IEEE Internet Things J.

Self Cite

View full text Add to dashboard Cite

Processing data generated at high volume and speed from the Internet of Things, smart cities, domotic, intelligent surveillance, and e-healthcare systems require efficient data processing and analytics services at the Edge to reduce the latency and response time of the applications. The Fog Computing Edge infrastructure consists of devices with limited computing, memory, and bandwidth resources, which challenge the construction of predictive analytics solutions that require resource-intensive tasks for training machine learning models. In this work, we focus on the development of predictive analytics for urban traffic. Our solution is based on deep learning techniques localized in the Edge, where computing devices have very limited computational resources. We present an innovative method for efficiently training of Gated Recurrent-Units (GRUs) across available resourceconstrained CPU and GPU Edge devices. Our solution employs distributed GRU model learning and dynamically stops the training process to utilize the low-power and resource-constrained Edge devices while ensuring good estimation accuracy effectively. The proposed solution was extensively evaluated using lowpowered ARM-based devices, including Raspberry Pi v3 and the low-powered GPU-enabled device NVIDIA Jetson Nano, and also compared them with Single-CPU Intel Xeon machines. For the evaluation experiments, we used real-world Floating Car Data. The experiments show that the proposed solution delivers excellent prediction accuracy and computational performance on the Edge when compared with the baseline methods.

show abstract

Section: B Comparison Of Single and Distributed Modelsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Automatic Distributed Deep Learning Using Resource-Constrained Edge Devices

Torre

Bahadori

Baig

et al. 2022

IEEE Internet Things J.

Self Cite

View full text Add to dashboard Cite

show abstract

“…Similarly, [26] attacks the problem of defining optimal thresholds for scaling policies with a reinforcement-learning algorithm that automatically and dynamically adjusts the thresholds without user configuration. Finally, [2] proposes an approach that uses a predictive autoscaling model trained on a dataset generated from simulations of reactive rule-based autoscaling. W.r.t.…”

Section: Related Work and Conclusionmentioning

confidence: 99%

“…W.r.t. work on workload prediction, such as [2], our global adaptation algorithm ability of detecting in advance service scaling needs is not based on guessing workload by means of logged data, but on mathematically calculating service MCL from system MCL (thanks to service Multiplicative Factor and current number of instances, see formula in Section 3.1). The two approaches are, thus, orthogonal: our approach avoids the negative consequences of the scaling chain effect, but it just passively waits for the triggering event (significant increment in the inbound workload).…”

Section: Related Work and Conclusionmentioning

confidence: 99%

Microservice Dynamic Architecture-Level Deployment Orchestration

Bacchiani

Bravetti

Giallorenzo

et al. 2021

Lecture Notes in Computer Science

View full text Add to dashboard Cite

We develop a novel approach for run-time global adaptation of microservice applications, based on synthesis of architecture-level reconfiguration orchestrations. More precisely, we devise an algorithm for automatic reconfiguration that reaches a target system Maximum Computational Load by performing optimal deployment orchestrations. To conceive and simulate our approach, we introduce a novel integrated timed architectural modeling/execution language based on an extension of the actor-based object-oriented Abstract Behavioral Specification (ABS) language. In particular, we realize a timed extension of SmartDeployer, whose ABS code annotations make it possible to express architectural properties. Our Timed SmartDeployer tool fully integrates time features of ABS and architectural annotations by generating timed deployment orchestrations. We evaluate the applicability of our approach on a realistic microservice application taken from the literature: an Email Pipeline Processing System. We prove its effectiveness by simulating such an application and by comparing architecture-level reconfiguration with traditional local scaling techniques (which detect scaling needs and enact replications at the level of single microservices). Our comparison results show that our approach avoids cascading slowdowns and consequent increased message loss and latency, which affect traditional local scaling.

show abstract

“…Abdullah et al 32 proposed an autoscaling model for requested services delivering on fog micro data centers to improve resource management and enhance response time. Their rule‐based model gathers the training dataset and builds the related prediction through preprocessing and postprocessing phases.…”

Section: Related Workmentioning

confidence: 99%

A proactive fog service provisioning framework for Internet of Things applications: An autonomic approach

Faraji‐MehmandarMohammad

Jabbehdari

Javadi

2021

Trans Emerging Tel Tech

View full text Add to dashboard Cite

In recent years, Internet of Things (IoT) services have expanded to promote the quality of life in different areas. Cloud connectivity services are so popular now that they have prompted the experts to enhance cloud computing for its utilization in IoT, making everything online in the next few decades. For reducing latency, immediate processing, and network congestion, fog computing has emerged in which cloud computing is expanded to the edge of the network. On the other hand, concerning the limitations in fog hardware resources compared with the cloud, and the dynamic and unpredictable fog environment, the provision of dynamic fog services is a challenge. Automatic matching of the resources based on the workload oscillations of IoT applications leads to allocating minimum fog resources to IoT devices, therefore, the satisfaction of service level agreement (SLA) and quality of service (QoS) parameters. The present article introduces a method based on the control monitoring‐analysis‐planning‐execution having shared knowledge‐base loop and presents an approach for dynamic resource provisioning based on autonomic computing and reinforcement learning techniques. The proposed scheme uses learning automata as a decision‐maker in the planning phase and time series prediction model in the analysis phase. The simulation test results indicated a reduced delay in service provisioning, total cost, and SLA violation compared with other approaches, highlighting the potential of fog computing in ensuring the QoS.

show abstract

Predictive Autoscaling of Microservices Hosted in Fog Microdata Center

Cited by 40 publications

References 44 publications

Automatic Distributed Deep Learning Using Resource-Constrained Edge Devices

Automatic Distributed Deep Learning Using Resource-Constrained Edge Devices

Microservice Dynamic Architecture-Level Deployment Orchestration

A proactive fog service provisioning framework for Internet of Things applications: An autonomic approach

Contact Info

Product

Resources

About