Title: Modeling and Analysis of Probabilistic Cloud Workflows and Their Applications
Ph.D. Candidate: Abdullah Alenizi
Major Advisor: Dr. Reda A. Ammar
Associate Advisors: Dr. Sanguthevar Rajasekaran, Dr. Song Han
Date/Time: Friday April 17, 2020 10:00 am-11:00 am
Meeting link: https://uconn-cmr.webex.com/uconn-cmr/j.php?MTID=m577dcca482058cc6613cfa291643f9d3
Meeting number: 619 030 584
Join by phone: +1-415-655-0002 US Toll
Cloud computing is increasingly becoming essential because of its flexibility and on-demand infrastructure it provides. As a result, new applications are targeting cloud infrastructure. Cloud computing has shifted the workload from local computers to centralized data centers which offer computing as a service. As a result, the cloud environment has become a great option for running cloud workflow applications. This type of application describes a job as a set of nodes and edges where the nodes represent tasks and the edges represent the dependencies between the tasks. Tasks can run in a sequence, parallel or choice patterns which are called Directed Acyclic Graph workflows. Non-Directed Acyclic Graph (non-DAG) allows iterations as well. Applications that include iteration and choice patterns may be stochastic. For such applications, execution time and dynamic cost depend on data being processed and can be determined during run-time.
In our research, we use the Computational Structure Model CSM to analyze cloud applications to understand more precise information about execution time and cost behavior. To conduct such a study, we provide a set of rules for transforming workflow graphs into equivalent CSM graphs. The generated model provides all possible execution times with their respective probabilities to generate more precise execution time estimates which can be used in better scheduling and resource allocation.
For a multi-job level, we introduce a queuing model system where jobs arrive at different times with different requirements (execution time, deadline,…etc). This new model will also help in utilizing the required resources and predicting the overall expected response time and its affiliated waiting time before submitting and during the execution of the jobs in the cloud. Having the execution and the waiting time in such a dynamic environment can help in evaluating the actual resources needed for submitted jobs to the cloud. Users can evaluate these metrics to decide if they will proceed at such time to complete the job or wait for another time to execute it. This gives a complete study and probabilistic behavior for a submitted job to the cloud. This study helps in determining performance metrics to compute different judgment of evaluating jobs and their time schedule.