Journal of Grid Computing最新文献

This paper analyzes the timing performance of a persistent storage designed for distributed container-based architectures in industrial control applications. The timing performance analysis is conducted using an in-house simulator, which mirrors our testbed specifications. The storage ensures data availability and consistency even in presence of faults. The analysis considers four aspects: 1. placement strategy, 2. design options, 3. data size, and 4. evaluation under faulty conditions. Experimental results considering the timing constraints in industrial applications indicate that the storage solution can meet critical deadlines, particularly under specific failure patterns. Comparison results also reveal that, while the method may underperform current centralized solutions in fault-free conditions, it outperforms the centralized solutions in failure scenario. Moreover, the used evaluation method is applicable for assessing other container-based critical applications with timing constraints that require persistent storage.

Recent modern computing and trends in digital transformation provide a smart healthcare system for predicting diseases at an early stage. In healthcare services, Internet of Things (IoT) based models play a vital role in enhancing data processing and detection. As IoT grows, processing data requires more space. Transferring the patient reports takes too much time and energy, which causes high latency and energy. To overcome this, Edge computing is the solution. The data is analysed in the edge layer to improve the utilization. This paper proposed effective prediction of resource allocation and prediction models using IoT and Edge, which are suitable for healthcare applications. The proposed system consists of three modules: data preprocessing using filtering approaches, Resource allocation using the Deep Q network, and prediction phase using an optimised DL model called DBN-LSTM with frog leap optimization. The DL model is trained using the training health dataset, and the target field is predicted. It has been tested using the sensed data from the IoT layer, and the patient health status is expected to take appropriate actions. With timely prediction using edge devices, doctors and patients conveniently take necessary actions. The primary objective of this system is to secure low latency by improving the quality of service (QoS) metrics such as makespan, ARU, LBL, TAT, and accuracy. The deep reinforcement learning approach is employed due to its considerable acceptance for resource allocation. Compared to the state-of-the-art approaches, the proposed system obtained reduced makespan by increasing the average resource utilization and load balancing, which is suitable for accurate real-time analysis of patient health status.

In different industries, there are miscellaneous applications that require multi-dimensional resources. These kinds of applications need all of the resource dimensions at the same time. Since the resources are typically scarce/expensive/pollutant, presenting an efficient resource allocation is a very favorable approach to reducing overall cost. On the other hand, the requirement of the applications on different dimensions of the resources is variable, usually, resource allocations have a high rate of wastage owing to the unpleasant resource skew-ness phenomenon. For instance, micro-service allocation in the Internet of Things (IoT) applications and Virtual Machine Placement (VMP) in a cloud context are challenging tasks because they diversely require imbalanced all resource dimensions such as CPU and Memory bandwidths, so inefficient resource allocation raises issues. In a special case, the problem under study associated with the two-dimensional resource allocation of distributed applications is modeled to the two-dimensional bin-packing problems which are categorized as the famous NP-Hard. Several approaches were proposed in the literature, but the majority of them are not aware of skew-ness and dimensional imbalances in the list of requested resources which incurs additional costs. To solve this combinatorial problem, a novel hybrid discrete gray wolf optimization algorithm (HD-GWO) is presented. It utilizes strong global search operators along with several novel walking-around procedures each of which is aware of resource dimensional skew-ness and explores discrete search space with efficient permutations. To verify HD-GWO, it was tested in miscellaneous conditions considering different correlation coefficients (CC) of resource dimensions. Simulation results prove that HD-GWO significantly outperforms other state-of-the-art in terms of relevant evaluation metrics along with a high potential of scalability.

In today’s smart environments, the serviceli-zation of various resources has produced a tremendous number of IoT- and cloud-based smart services. Thanks to the pivotal role of pillar paradigms, such as edge/cloud computing, Internet of Things, and business process management, it is now possible to combine and translate these service-like resources into configurable workflows, to cope with users’ complex needs. Examples include treatment workflows in smart healthcare, delivery plans in drone-based missions, transportation plans in smart urban networks, etc. Rather than composing atomic services to obtain these workflows, reusing existing process fragments has several advantages, mainly the fast, secure, and configurable compositions. However, reusing smart process fragments has not yet been addressed in the context of smart environments. In addition, existing solutions in smart environments suffer from the complexity (e.g., multi-modal transportation in smart mobility) and privacy issues caused by the heterogeneity (e.g., package delivery in smart economy) of aggregated services. Moreover, these services may be conflicting in specific domains (e.g. medication/treatment workflows in smart healthcare), and may affect user experience. To solve the above issues, the present paper aims to accelerate the process of generating configurable treatment workflows w.r.t. the users’ requirements and their smart environment specificity. We exploit the principles of software reuse to map each sub-request into smart process fragments, which we combine using Cocke-Kasami-Younger (CKY) method, to finally obtain the suitable workflow. This contribution is preceded by a knowledge graph modeling of smart environments in terms of available services, process fragments, as well as their dependencies. The built information network is, then, managed using a graph representation learning method, in order to facilitate its processing and composing high-quality smart services. Experimental results on a real-world dataset proved the effectiveness of our approach, compared to existing solutions.