Modern big data applications tend to prefer a cluster computing approach as they are linked to the distributed computing framework that serves users jobs as per demand. It performs rapid processing of tasks by subdividing them into tasks that execute in parallel. Because of the complex environment, hardware and software issues, tasks might run slowly leading to delayed job completion, and such phenomena are also known as stragglers. The performance improvement of distributed computing framework is a bottleneck by straggling nodes due to various factors like shared resources, heavy system load, or hardware issues leading to the prolonged job execution time. Many state-of-the-art approaches use independent models per node and workload. With increased nodes and workloads, the number of models would increase, and even with large numbers of nodes. Not every node would be able to capture the stragglers as there might not be sufficient training data available of straggler patterns, yielding suboptimal straggler prediction. To alleviate such problems, we propose a novel collaborative learning-based approach for straggler prediction, the alternate direction method of multipliers (ADMM), which is resource-efficient and learns how to efficiently deal with mitigating stragglers without moving data to a centralized location. The proposed framework shares information among the various models, allowing us to use larger training data and bring training time down by avoiding data transfer. We rigorously evaluate the proposed method on various datasets with high accuracy results.

/pdf/collaborative-learning-based-straggler-prevention-in-large-3s319b5eev.pdf

Collaborative Learning Based Straggler Prevention in Large-Scale Distributed Computing Framework

Modified Social Group Optimization-a meta-heuristic algorithm to solve short-term hydrothermal scheduling

Abstract Carcinoma is a primary source of morbidity in women globally, with metastatic disease accounting for most deaths. Its early discovery and diagnosis may significantly increase the odds of survival. Breast cancer imaging is critical for early identification, clinical staging, management choices, and treatment planning. In the current study, the FastAI technology is used with the ResNet-32 model to precisely identify ductal carcinoma. ResNet-32 is having few layers comparted to majority of its counterparts with almost identical performance. FastAI offers a rapid approximation toward the outcome for deep learning models via GPU acceleration and a faster callback mechanism, which would result in faster execution of the model with lesser code and yield better precision in classifying the tissue slides. Residual Network (ResNet) is proven to handle the vanishing gradient and effective feature learning better. Integration of two computationally efficient technologies has yielded a precision accuracy with reasonable computational efforts. The proposed model has shown considerable efficiency in the evaluating parameters like sensitivity, specificity, accuracy, and F1 Score against the other dominantly used deep learning models. These insights have shown that the proposed approach might assist practitioners in analyzing Breast Cancer (BC) cases appropriately, perhaps saving future complications and death. Clinical and pathological analysis and predictive accuracy have been improved with digital image processing. 

/pdf/resnet-32-and-fastai-for-diagnoses-of-ductal-carcinoma-from-31i3a05q.pdf

ResNet-32 and FastAI for diagnoses of ductal carcinoma from 2D tissue slides

With the rapid increase in the use of cloud computing systems, an efficient task scheduling policy, which deals with the assignment of tasks to resources, is required to obtain maximum performance. Cloud task scheduling (CTS) is an established NP-Hard optimization problem that can be effectively tackled with meta-heuristic algorithms. The cuckoo search (CS) algorithm is a powerful swarm-intelligence meta-heuristic that has been successfully applied over a wide-range of real-life optimization problems, including task scheduling problems. Besides its strong exploration ability, the CS algorithm suffers from insufficient local search, lack of solution diversity towards the end, and slow convergence problem. These drawbacks produce inefficient cloud task schedules resulting in sub-optimal performance. In this manuscript, an improved CS-based scheduling algorithm called CSDEO is introduced, which combines the features of the Opposition-based learning (OBL) method, Cuckoo search, and Differential evolution (DE) algorithms to optimize workload makespan and energy consumption of the cloud resources. Our CSDEO algorithm firstly uses the OBL method to produce an optimal initial population by providing solutions across the entire solution space. Then, the CSDEO uses an effective way of switching between the CS exploration phase and the DE exploitation phase, depending on each solution's fitness. Experiments are conducted on the CloudSim simulator by using the CEA-Curie and HPC2N supercomputing workloads. The observations show that in the case of CEA-Curie workloads, the proposed CSDEO algorithm achieves makespan improvement in the range of 6.29–29.76% and energy consumption improvement in the range of 3.76–201.98% over well-known scheduling algorithms. In the case of HPC2N workloads, the improvement ranges of the CSDEO approach for the makespan and energy consumption metrics are 9.86–281.69% and 6.12–233.3%, respectively compared to the tested scheduling algorithms.

Multi-criteria HPC task scheduling on IaaS cloud infrastructures using meta-heuristics

Scheduling is a considerable problem in cloud to increase the quality of service provisioning with higher resource efficiency. The conventional task scheduling algorithms designed for balancing load in a cloud environment. But, minimizing the Service level agreement (SLA) violation, resource wastage and the energy consumption during the task scheduling process was not solved effectively. In order to resolve these limitations, a new virtual machine (VM) consolidation technique called Nature-inspired Meta-heuristic Threshold based firefly optimized lottery scheduling (NMT-FOLS) Technique is proposed. Initially, user requests are transmitted to the cloud server (CS). Next, NMT-FOLS Technique utilizes Adaptive Regressive Holt–Winters Workload Predictor to discover the workload state as normal or timely or bursty. Using the workload predictor result, NMT-FOLS Technique exploits task scheduler to allocate user requested tasks to optimal VMs. NMT-FOLS Technique applies multi-objective firefly optimization based task scheduling algorithm in normal workload state and multi-objective firefly optimized lottery scheduling algorithm in timely and bursty workload situations. At last, the selected scheduling algorithm in NMT-FOLS Technique assigns the user requested task to best VMs in CS to perform the demanded services. Hence, NMT-FOLS Technique gets better task scheduling performance to balance normal, timely and bursty workloads in CS with lesser time. NMT-FOLS Technique decreases the SLA violation in cloud through scheduling of user tasks to optimal VM. NMT-FOLS performs an experimental process using metrics such as SLA violation, task scheduling efficiency (TSE), makespan, energy utilization and memory usage with number of user-requested tasks over the considered Amazon dataset. From the experimental result, the NMT-FOLS technique improves scheduling efficiency up to 94.6% and reduces the SLA violations and energy utilization from different test cases on an average to 78%, and 63% compared to state-of-the-art works.

Adaptive regressive holt–winters workload prediction and firefly optimized lottery scheduling for load balancing in cloud

Effective resource distribution to regulate load uniformly in heterogeneous cloud environments is crucial. Resource allotment which is taken after capable task scheduling is a critical worry in cloud environment. The incoming job requests are assigned to resources equally by load balancer in such a way that resources are utilized effectively. Number of cloud clients is very great in number, degree of approaching job requests is uninformed and information is tremendous in cloud application. Resources in cloud environment are constrained. Hence, it is not easy to deploy different applications with unpredictable limits and functionalities in heterogeneous cloud environment. The proposed method has two phases such as allocation of resources and scheduling of tasks. Effective resource allocation is proposed using social group optimization algorithm and scheduling of tasks using shortest-job-first scheduling algorithm for minimizing the makespan time and maximizing throughput. Experimentations are performed for accurate simulations on artificial data for heterogeneous cloud environment. Experimental results are compared with first-in, first-out and genetic algorithm-based shortest-job-first scheduling. Validity of the proposed method noticeably gives improved performance of the system in provisions of makespan time and throughput.

Effective Allocation of Resources and Task Scheduling in Cloud Environment using Social Group Optimization

The adequate aging hypothesis seeks to help people live longer, healthy lives. Diabetic patients who stay remotely need an infrastructure to monitor them continuously and provide timely treatment. Ambient assisted living (AAL) encourages the establishment of solutions that may help optimize older people’s assistive environment while also reducing their impairments. The blood glucose levels of diabetic patients are continuously monitored by gold oxide sensors placed over the human body. The signals associated with the glucose levels in the human body are plotted over a spectrogram image using the short-time Fourier transform, which is further classified using the deep learning model based on finetuned AlexNet, which has employed random oversampling and batch normalization for better precision in the results. The model classifies the spectrogram images as low and high glucose levels and normal glucose levels. Thereby alarming the caretakers for effective treatment of the individuals. Body area networks (BANs) gather information from biosensors and send it to a domain controller to assist caretakers and physicians in recommending the physical exercises for their clients. Evaluation criteria such as sensitivity and specificity, precision, and Mathew’s correlation coefficient are used to assess the effectiveness of the proposed model in this current diabetes study. The cross-validation of the model at multiple folds is being evaluated to analyze the performance. It is evident from the obtained results that the proposed model has exhibited an acceptable performance in precisely sensing the individuals with abnormal glucose levels.

Ambient Assistive Living for Monitoring the Physical Activity of Diabetic Adults through Body Area Networks

Confidentiality and security breaches remain significant concerns for electronic healthcare systems even though many relevant rules, principles, and compliance standards are in place to protect health information. It is now essential to use strategies revolving around big data to enhance the dependability of healthcare delivery due to the ever-increasing number of data generated within the healthcare sector. Even though big data processing methods and platforms have been incorporated into the data management designs for medical systems, these designs have trouble addressing urgent situations. Today, it is difficult to predict how big data and ML will affect the healthcare sectors. As a result, a clinical healthcare data warehouse environment utilizing big data analytics and ML is provided in this analysis. Quick digital access to all types of vital data, including patient histories, scan records, insurance claims, and payment history, is guaranteed by healthcare data warehouse systems. This clinical healthcare data warehouse environment can improve an individual's quality of care and monitor the patient's health status in real-time by using ML algorithms and Big data analytics. The effectiveness of the ML technique's performance is measured in terms of accuracy, specificity, sensitivity, precision, recall and Receiver Operating Characteristics (RoC). 

A robust framework for handling health care information based on machine learning and big data engineering techniques

Often cloud providers and cloud clients illustrate several constraints and thus allocation of resources in a heterogeneous cloud is a difficult job. As the traffic flow is quite subjective and Client necessities and applications size vary regularly, the major challenge and concern is to map the external job requests to available virtual machines. To reduce the gap among regularly altering client requirements and existing resources, Client-Awareness Allocation of Resources and Scheduling of jobs in cloud by using social group optimization (SGOCARAJS) is proposed. This algorithm is mainly split into two phases namely allocation of resources using SGO and shortest job first scheduling. The main aim is to map the jobs to virtual machines of cloud group to attain higher client satisfaction and lowest makespan time. Experiments are conducted on datasets and results are compared with present scheduling techniques. This model proved that this algorithm outrun the available algorithms based on concerned metrics.

S. Phani Praveen

Papers

Effective Allocation of Resources and Task Scheduling in Cloud Environment using Social Group Optimization

Ambient Assistive Living for Monitoring the Physical Activity of Diabetic Adults through Body Area Networks

ResNet-32 and FastAI for diagnoses of ductal carcinoma from 2D tissue slides

A robust framework for handling health care information based on machine learning and big data engineering techniques

Client-Awareness Resource Allotment and Job Scheduling in Heterogeneous Cloud by Using Social Group Optimization