Joint‐analysis of performance and energy consumption when enabling cloud elasticity for synchronous HPC applications

Abstract

Summary A key characteristic of cloud computing is elasticity, automatically adjusting system resources to an application's workload. Both reactive and horizontal approaches represent traditional means to offer this capability, in which rule‐condition‐action statements and upper and lower thresholds occur to instantiate or consolidate compute nodes and virtual machines. Although elasticity can be beneficial for many HPC (high‐performance computing) scenarios, it also imposes significant challenges in the development of applications. In addition to issues related to how we can incorporate this new feature in such applications, there is a problem associated with the performance and resource pair and, consequently, with energy consumption. Further exploring this last difficulty, we must be capable of analyzing elasticity effectiveness as a function of employed thresholds with clear metrics to compare elastic and non‐elastic executions properly. In this context, this article explores elasticity metrics in two ways: (i) the use of a cost function that combines application time with different energy models; (ii) the extension of speedup and efficiency metrics, commonly used to evaluate parallel systems, to cover cloud elasticity. To accomplish (i) and (ii), we developed an elasticity model known as AutoElastic, which reorganizes resources automatically across synchronous parallel applications. The results, obtained with the AutoElastic prototype using the OpenNebula middleware, are encouraging. Considering a CPU‐bound application, an upper threshold close to 70% was the best option for obtaining good performance with a non‐prohibitive elasticity cost. In addition, the value of 90% for this threshold was the best option when we plan an efficiency‐driven execution. Copyright © 2015 John Wiley & Sons, Ltd.

Towards Evaluating Proactive and Reactive Approaches on Reorganizing Human Resources in IoT-Based Smart Hospitals

“…So, based in Brazilian Law Project of 14 June 2018 [55] that proposes a maximum waiting time for care in hospitals, clinics, and laboratories of 30 min on regular days (from Monday to Sunday), we define ElHealth’s maximum load (i.e., 100%) in 30 min. Based on several works [12,47,56,57], we are using 4 combinations of thresholds when evaluating the second scenario, so considering 30% (9 min) and 50% (15 min) for lower threshold, and considering 70% (21 min) and 90% (27 min) for upper threshold. For proactive elasticity, we set ElHealth’s upper threshold in 30 min, (i.e., maximum load previously defined), and we set ElHealth’s lower threshold in 9 min (30% of maximum waiting time).…”

“…However, this value refers to a hospital room with a single attendant allocated for care, but in most cases will be more than one health professional working in that room, making it necessary to identify the average time with different numbers of attendants. In this context, ElHealth model uses a parallel allocation of human resources, such as the parallel allocation of virtual machines used in elastic systems [13] or the use of parallel processors in high-performance computing [47]. Thus, based on the Elastic Speedup proposed by [47], ElHealth introduces Equation (5) for Human Resources Elastic Speedup.…”

“…In this context, ElHealth model uses a parallel allocation of human resources, such as the parallel allocation of virtual machines used in elastic systems [13] or the use of parallel processors in high-performance computing [47]. Thus, based on the Elastic Speedup proposed by [47], ElHealth introduces Equation (5) for Human Resources Elastic Speedup. Considering again the previous example (Figure 9), with room r1 between two times ti and tf with an average number of 4 patients, an average care time of 10 min and with two health professionals allocated, as shown in Figure 10.…”

Multi-Hospital Management: Combining Vital Signs IoT Data and the Elasticity Technique to Support Healthcare 4.0

“…Consequently, when our model detects waiting times that are not in accordance with established limits, HealCity must calculate the amount of health resources required to meet the patient's demand, thus recognizing the need for adjustments in that particular room. HealCity model uses a parallel allocation of human resources, inspired by similar strategies used in elastic systems [38] and high-performance computing [44]. Therefore, HealCity introduces a mathematical formalism to calculate the Proactive Human Resources Elastic Speedup (PHRES), details of which will follow.…”

FPGA-Aware Scheduling Strategies at Hypervisor Level in Cloud Environments

“…The first one describes the behavior of the whole system. It is composed of the total energy used for the system to process a bag of tasks (energy-to-solution [31]). The second group is relative to SLA compliance, the percentage of SLA fulfilled successfully, and the average energy invested per fulfilled SLA.…”