Utility-based Scheduling Solution for Scientific Workflow on Cloud

Data-intensive scientific applications are posing many challenges in distributed computing systems. In the scientific field, the application data are expected to double every year over the next decade and further. With this continuing data explosion, high performance computing systems are needed to store and process data efficiently, and workflow technologies are facilitated to automate these scientific applications. Scientific workflows are typically very complex. They usually have a large number of tasks and need a long time for execution. Running scientific workflow applications usually need not only high performance computing resources but also massive storage. The emergence of cloud computing technologies offers a new way to develop scientific workflow systems. Scientists can upload their data and launch their applications on the scientific cloud workflow systems from everywhere in the world via the Internet, and they only need to pay for the resources that they use for their applications. As all the data are managed in the cloud, it is easy to share data among scientists. This kind of model is very convenient for users, but remains a big challenge to the system. This paper proposes several research topics of data management in scientific cloud workflow systems, and discusses their research methodologies and state-of-the-art solutions.

Scientific workflows are one of the well-established pillars of modern large-scale computational science. More recently, scientists have started to leverage machine learning (ML) capabilities in their workflows, leading to a new category of scientific workflows, denoted as scientific ML workflows. ML is not only about training and inference, modern ML workflows also involve complex data processing steps before the training can start, which are not often accounted for in most performance studies. In this work, we consider scientific ML workflows, from data pre-processing to training, inference, and model evaluation. We aim to explore (i) how scientific ML workflows differ from more traditional scientific workflows and; (ii) how we can characterize ML workflows both in terms of execution time and data movements when executing on a target cloud platform. We select three representative workflows, ranging from image classification to natural language processing and image segmentation, which have been executed using the academic cloud platform, Chameleon. We build four realistic deployment scenarios for each workflow, which stress data movements during workflow executions. Then, we compare the performance observed when utilizing these different configurations and study how different settings impact overall workflows performance and efficiency when running on cloud infrastructures. Finally, we summarize our findings and discuss performance impacts when augmenting scientific workflows with ML techniques and how traditional workflow management systems can improve their support for such workflows.

The public cloud environment has emerged as a promising platform for exe-cuting scientific workflows. These executions involve leasing virtual machines (VMs) from public services for the duration of the workflow. The structure of the workflows significantly impacts the performance of any proposed scheduling approach. A task within a workflow cannot begin its execution before receiving all the required data from its preceding tasks. In this paper, we introduce a multi-priority scheduling approach for executing workflow tasks in the cloud. The key component of the proposed approach is a mechanism that logically or-ders and groups workflow tasks based on their data dependencies and locality. Using the proposed approach, the number of available VMs influences the num-ber of groups (partitions) obtained. Based on the locality of each group’s tasks, the priority of each group is determined to reduce the overall execution delay and improve VM utilization. As the results demonstrate, the proposed approach achieves a significant reduction in both execution costs and time in most scenar-ios

Executing scientific workflows in high-performance cloud computing platforms requires the use of scheduling algorithms that allow workflows execution as fast as possible, while minimizing the monetary cost of such executions. In this work we study how the use of supporting information can offer guidance to scheduling algorithms, helping them to devise more efficient execution plans in terms of the total execution time (makespan) and the total monetary cost. Using two large-scale scientific workflows, our experiments showed that simple modifications on a classical scheduling algorithm (HEFT), in conjunction with the appropriate supporting information, could reduce the monetary cost of an execution in up to 59% and reduce the makespan in up to 8.6%.

Moldable scientific workflows represent a special class of scientific workflows where the tasks are written as distributed programs being able to exploit various amounts of computer resources. However, current cluster job schedulers require the user to specify the amount of resources per task manually. This often leads to suboptimal execution time and related cost of the whole workflow execution since many users have only limited experience and knowledge of the parallel efficiency and scaling. This paper proposes several mechanisms to automatically optimize the execution parameters of moldable workflows using genetic algorithms. The paper introduces a local optimization of workflow tasks, a global optimization of the workflow on systems with on-demand resource allocation, and a global optimization for systems with static resource allocation. Several objectives including the workflow makespan, computational cost and the percentage of idling nodes are investigated together with a trade-off parameter putting stress on one objective or another. The paper also discusses the structure and quality of several evolved workflow schedules and the possible reduction in makespan or cost. Finally, the computational requirements of evolutionary process together with the recommended genetic algorithm settings are investigated. The most complex workflows may be evolved in less than two minutes using the global optimization while in only 14 s using the local optimization.

This paper presents a cost optimization model for scheduling scientific workflows on IaaS clouds such as Amazon EC2 or RackSpace. We assume multiple IaaS clouds with heterogeneous virtual machine instances, with limited number of instances per cloud and hourly billing. Input and output data are stored on a cloud object store such as Amazon S3. Applications are scientific workflows modeled as DAGs as in the Pegasus Workflow Management System. We assume that tasks in the workflows are grouped into levels of identical tasks. Our model is specified using mathematical programming languages (AMPL and CMPL) and allows us to minimize the cost of workflow execution under deadline constraints. We present results obtained using our model and the benchmark workflows representing real scientific applications in a variety of domains. The data used for evaluation come from the synthetic workflows and from general purpose cloud benchmarks, as well as from the data measured in our own experiments with Montage, an astronomical application, executed on Amazon EC2 cloud. We indicate how this model can be used for scenarios that require resource planning for scientific workflows and their ensembles.

With the rapid development of processing and storage technologies and the success of the Internet, computing resources have become cheaper, more powerful and more ubiquitously available than ever before. This technological trend has enabled the realization of a new computing model, called cloud computing. In cloud, scheduling is an important application. In cloud environments, load balancing task scheduling is an important problem that directly affects resource utilization. Undoubtedly, load balancing scheduling is a serious aspect that should be considered because of its significant impact on both the back end and the front end of the cloud research industry. Good resource utilization is achieved whenever an effective load balance is achieved in the cloud. But, load balancing in cloud computing is an NP-hard optimization problem. In order to accomplish this problem, a novel load balancing task scheduling algorithm in cloud using Adaptive Dragonfly algorithm (ADA) is proposed. The ADA is a combination of dragonfly algorithm and firefly algorithm. Moreover, to attain the better performance, multi-objective function is developed based on three parameters namely, completion time, processing costs and load. Finally, the performance of proposed methodology is evaluated in terms of different metrics namely, execution cost and execution time. The experimental results demonstrate that a proposed approach accomplishes better load balancing result compared to other approaches.

Scheduling large-scale and resource-intensive workflows in cloud infrastructure is one of the main challenges for cloud service providers (CSPs). Cloud infrastructure is more efficient when virtual machines and other resources work up to their full potential. The main factor that influences the quality of cloud services is the distribution of workflow on virtual machines (VMs). Scheduling tasks to VMs depends on the type of workflow and mechanism of resource allocation. Scientific workflows include large-scale data transfer and consume intensive resources of cloud infrastructures. Therefore, scheduling of tasks from scientific workflows on VMs requires efficient and optimized workflow scheduling techniques. This paper proposes an optimised workflow scheduling approach that aims to improve the utilization of cloud resources without increasing execution time and execution cost.

Fog computing aims to provide the solution of bandwidth, network latency and energy consumption problems of cloud computing. Likewise, management of data generated by healthcare IoT devices is one of the significant applications of fog computing. Huge amount of data is being generated by healthcare IoT devices and such types of data is required to be managed efficiently, with low latency, without failure, and with minimum energy consumption and low cost. Failures of task or node can cause more latency, maximum energy consumption and high cost. Thus, a failure free, cost efficient, and energy aware management and scheduling scheme for data generated by healthcare IoT devices not only improves the performance of the system but also saves the precious lives of patients because of due to minimum latency and provision of fault tolerance. Therefore, to address all such challenges with regard to data management and fault tolerance, we have presented a Fault Tolerant Data management (FTDM) scheme for healthcare IoT in fog computing. In FTDM, the data generated by healthcare IoT devices is efficiently organized and managed through well-defined components and steps. A two way fault-tolerant mechanism i.e., task-based fault-tolerance and node-based fault-tolerance, is provided in FTDM through which failure of tasks and nodes are managed. The paper considers energy consumption, execution cost, network usage, latency, and execution time as performance evaluation parameters. The simulation results show significantly improvements which are performed using iFogSim. Further, the simulation results show that the proposed FTDM strategy reduces energy consumption 3.97%, execution cost 5.09%, network usage 25.88%, latency 44.15% and execution time 48.89% as compared with existing Greedy Knapsack Scheduling (GKS) strategy. Moreover, it is worthwhile to mention that sometimes the patients are required to be treated remotely due to non-availability of facilities or due to some infectious diseases such as COVID-19. Thus, in such circumstances, the proposed strategy is significantly efficient.

Data reduction in scientific workflows using provenance monitoring and user steering

Most of the large-scale scientific experiments modeled as scientific workflows produce a large amount of data and require workflow parallelism to reduce workflow execution time. Some of the existing Scientific Workflow Management Systems (SWfMS) explore parallelism techniques - such as parameter sweep and data fragmentation. In those systems, several computing resources are used to accomplish many computational tasks in homogeneous environments, such as multiprocessor machines or cluster systems. Cloud computing has become a popular high performance computing model in which (virtualized) resources are provided as services over the Web. Some scientists are starting to adopt the cloud model in scientific domains and are moving their scientific workflows (programs and data) from local environments to the cloud. Nevertheless, it is still difficult for the scientist to express a parallel computing paradigm for the workflow on the cloud. Capturing distributed provenance data at the cloud is also an issue. Existing approaches for executing scientific workflows using parallel processing are mainly focused on homogeneous environments whereas, in the cloud, the scientist has to manage new aspects such as initialization of virtualized instances, scheduling over different cloud environments, impact of data transferring and management of instance images. In this paper we propose SciCumulus, a cloud middleware that explores parameter sweep and data fragmentation parallelism in scientific workflow activities (with provenance support). It works between the SWfMS and the cloud. SciCumulus is designed considering cloud specificities. We have evaluated our approach by executing simulated experiments to analyze the overhead imposed by clouds on the workflow execution time.

Cloud computing is recently a booming area and has been emerging as a commercial reality in the information technology domain. Cloud computing represents supplement, consumption and delivery model for IT services that are based on internet on pay as per usage basis. The scheduling of the cloud services to the consumers by service providers influences the cost benefit of this computing paradigm. In such a scenario, Tasks should be scheduled efficiently such that the execution cost and time can be reduced. In this paper, we proposed a meta-heuristic based scheduling, which minimizes execution time and execution cost as well. An improved genetic algorithm is developed by merging two existing scheduling algorithms for scheduling tasks taking into consideration their computational complexity and computing capacity of processing elements. Experimental results show that, under the heavy loads, the proposed algorithm exhibits a good performance.

Cloud computing is designed for multiple users who can access services, accommodate large volumes of data and build applications. Managing all the resources and to provide on-demand access among several users is complex and difficult. All the inter-dependent tasks need to be properly scheduled and managed so that the resources are utilized effectively and tasks are executed on a lesser time with low cost. Several scheduling algorithms were proposed to schedule tasks and satisfy objectives such as either cost improvement or execution time improvement or server utilization. To improve the scheduling efficiency and to obtain multiple objectives like reduced execution time, cost and server utilization a framework for effective scheduling of tasks in cloud environment is proposed based on the Artificial Bee Colony (ABC) The proposed framework is analyzed in terms of execution time, cost and server utilization and the results are optimal for an effective cloud environment.

In mobile wireless multimedia sensor networks' image compression task collaborations, the existing methods do not consider the dynamic changes in the processing ability and the locations of the cooperative nodes. when processing the image compression tasks, these methods will cause frequent interruptions and result in data re-transmission for those tasks. in this paper, an image compression task collaboration algorithm based on dynamic alliance is proposed to solve this problem in mobile wireless multimedia sensor networks. first, a dynamic task alliance is established by the camera node based on the location, computing capability and resource usage of ordinary nodes. then, the location and average moving velocity of the camera nodes and ordinary nodes are considered to calculate the task stable execution time. the image compression task is divided into an image transfer sub task and an image compression sub task based on the task stable execution time. finally, an image compression task collaboration allocation optimization model is established according to the transmission time, the execution time, the execution cost, and the network energy consumption. the gradient method is used to realize the cooperative allocation of image compression tasks. simulation results show that the proposed algorithm can realize task load balancing, reduce the execution time, and network energy consumption.

The ability to accurately estimate the execution time of computationally expensive e-science algorithms enables better scheduling of workflows that incorporate those algorithms as their building blocks, and may give users an insight into the expected cost of workflow execution on cloud resources. When a large history of past runs can be observed, crude estimates such as the average execution time can easily be provided. We make the hypothesis that, for some algorithms, better estimates can be obtained by using the histories to learn regression models that predict execution time based on selected features of their inputs. We refer to this property as input predictability of algorithms. We are motivated by e-science workflows that involve repetitive training of multiple learning models. Thus, we verify our hypothesis on the specific case of the C4.5 decision tree builder, a well-known learning method whose training execution time is indeed sensitive to the specific input dataset, but in non- obvious ways. We use the case study to demonstrate a method for assessing input predictability. While this yields promising results, we also find that its more general applicability involves a trade off between the black-box nature of the algorithms under analysis, and the need for expert insight into relevant features of their inputs.