Workflow Performance Analysis Research Articles

Computational reproducibility of scientific workflows at extreme scales

We propose an approach for improved reproducibility that includes capturing and relating provenance characteristics and performance metrics. We discuss two use cases: scientific reproducibility of results in the Energy Exascale Earth System Model (E3SM—previously ACME) and performance reproducibility in molecular dynamics workflows on HPC platforms. To capture and persist the provenance and performance data of these workflows, we have designed and developed the Chimbuko and ProvEn frameworks. Chimbuko captures provenance and enables detailed single workflow performance analysis. ProvEn is a hybrid, queryable system for storing and analyzing the provenance and performance metrics of multiple runs in workflow performance analysis campaigns. Workflow provenance and performance data output from Chimbuko can be visualized in a dynamic, multilevel visualization providing overview and zoom-in capabilities for areas of interest. Provenance and related performance data ingested into ProvEn is queryable and can be used to reproduce runs. Our provenance-based approach highlights challenges in extracting information and gaps in the information collected. It is agnostic to the type of provenance data it captures so that both the reproducibility of scientific results and that of performance can be explored with our tools.

Workflow performance analysis and simulation based on multidimensional workflow net

Workflow model performance analysis plays an important role in the research of workflow techniques and efficient implementation of workflow management. Instances dwelling times (IDT) which consist of waiting times and handle times in a workflow model is a key performance analysis goal. In a workflow model the instances which act as customers and the resources which act as servers form a queuing network. Multidimensional workflow net (MWF-net) includes multiple timing workflow nets (TWF-nets) and the organization and resource information. This paper uses queuing theory and MWF-net to discuss mean value and probability distribution density function (PDDF) of IDT. It is assumed that the instances arrive with exponentially distributed inter-arrival times and the resources handle instances within exponentially distributed times or within constant times. First of all, the mean value and PDDF of IDT in each activity is calculated. Then the mean value and PDDF of IDT in each control structure of a workflow model is computed. According to the above results a method is proposed for computing the mean value and PDDF of IDT in a workflow model. Finally an example is used to show that the proposed method can be effectively utilized in practice.

Calculating the Maximum Throughput in the Context of Workflow Execution

Workflow performance analysis is the basis of workflow evaluation and optimization. Maximum throughput is an important aspect of workflow performance. In this paper, a process analytic model for workflow execution is firstly presented. Then an algorithm is developed to identify the bottleneck activity and to calculate maximum throughput, which is a good preparation for the further analysis and optimization of workflow. Finally the simulation results show the accuracy of the proposed algorithm.