WfInstances: Workflow Instances

Workflow execution instances have been widely used to profile and characterize workflow executions, and to build distributions of workflow execution behaviors, which are used to evaluate methods and techniques in simulation or in real conditions.

The WfCommons project targets the analysis of actual workflow execution instances (i.e., the workflow execution profile data and characterizations) in order to build Workflow Recipes of workflow applications. These recipes contain the necessary information for generating synthetic, yet realistic, workflow instances that resemble the structure and distribution of the original workflow executions.

A list of workflow execution instances that are compatible with WfFormat is kept constantly updated in our project website.

WfInstances

A workflow execution instance represents an actual execution of a scientific workflow on a distributed platform (e.g., clouds, grids, HPC, etc.). In the WfCommons project, an instance is represented in a JSON file following the schema described in WfFormat. This Python package provides an instance loader tool for importing workflow execution instances for analysis. For instance, the code snippet below shows how an instance can be loaded using the Instance class:

import pathlib
from wfcommons import Instance
input_instance = pathlib.Path('/path/to/instance/file.json')
instance = Instance(input_instance=input_instance)

The Instance class provides a number of methods for interacting with the workflow instance, including:

draw(): produces an image or a pdf file representing the instance.
leaves(): gets the leaves of the workflow (i.e., the tasks without any successors).
roots(): gets the roots of the workflow (i.e., the tasks without any predecessors).
write_dot(): writes a dot file of the instance.

Note

Although the analysis methods are inherently used by WfCommons (specifically WfChef) for WfChef: Workflows Recipes, they can also be used in a standalone manner.

Parsing Workflow Execution Logs

The most common way for obtaining workflow instances from actual workflow executions is to parse execution logs. As part of the WfCommons project, we are constantly developing parsers for commonly used workflow management systems. The parsers provided in this Python package automatically scans execution logs to produce instances using WfFormat.

Each parser class is derived from the abstract LogsParser class. Thus, each parser provides a build_workflow() method.

Makeflow

Makeflow is a workflow system for executing large complex workflows on clusters, clouds, and grids. The Makeflow language is similar to traditional “Make”, so if you can write a Makefile, then you can write a Makeflow. A workflow can be just a few commands chained together, or it can be a complex application consisting of thousands of tasks. It can have an arbitrary DAG structure and is not limited to specific patterns. Makeflow is used on a daily basis to execute complex scientific applications in fields such as data mining, high energy physics, image processing, and bioinformatics. It has run on campus clusters, the Open Science Grid, NSF XSEDE machines, NCSA Blue Waters, and Amazon Web Services. Makeflow logs provide time-stamped event instances from these executions. The following example shows the analysis of Makeflow execution logs, stored in a local folder (execution_dir), for a workflow execution using the MakeflowLogsParser class:

import pathlib
from wfcommons.wfinstances import MakeflowLogsParser

# creating the parser for the Makeflow workflow
execution_dir = pathlib.Path('/path/to/makeflow/execution/dir/blast/chameleon-small-001/')
resource_monitor_logs_dir = pathlib.Path('/path/to/makeflow/resource/monitor/logs/dir')
parser = MakeflowLogsParser(execution_dir=execution_dir,
                            resource_monitor_logs_dir=resource_monitor_logs_dir)

# generating the workflow instance object
workflow = parser.build_workflow('makeflow-workflow-test')

# writing the workflow instance to a JSON file
workflow_path = pathlib.Path('./makeflow-workflow.json')
workflow.write_json(workflow_path)

Note

The MakeflowLogsParser class requires that Makeflow workflows to run with the Resource Monitor tool (e.g., execute the workflow using the --monitor=logs).

Nextflow

Nextflow is a reactive workflow framework and a programming DSL that eases the writing of data-intensive computational pipelines. It is designed around the idea that the Linux platform is the lingua franca of data science. Linux provides many simple but powerful command-line and scripting tools that, when chained together, facilitate complex data manipulations. Nextflow extends this approach, adding the ability to define complex program interactions and a high-level parallel computational environment based on the dataflow programming model. The following example shows the analysis of Nextflow execution logs, stored in a local folder (execution_dir), for a workflow execution using the NextflowLogsParser class:

import pathlib
from wfcommons.wfinstances import NextflowLogsParser

# creating the parser for the Nextflow workflow
execution_dir = pathlib.Path('/path/to/nextflow/execution/dir/')
parser = NextflowLogsParser(execution_dir=execution_dir)

# generating the workflow instance object
workflow = parser.build_workflow('nextflow-workflow-test')

# writing the workflow instance to a JSON file
workflow_path = pathlib.Path('./nextflow-workflow.json')
workflow.write_json(workflow_path)

Note

The NextflowLogsParser class assumes that workflow executions will produce an execution_report_*.html and an execution_timeline_*.html files.

Pegasus

Pegasus is being used in production to execute workflows for dozens of high-profile applications in a wide range of scientific domains. Pegasus provides the necessary abstractions for scientists to create workflows and allows for transparent execution of these workflows on a range of compute platforms including clusters, clouds, and national cyberinfrastructures. Workflow execution with Pegasus includes data management, monitoring, and failure handling, and is managed by HTCondor DAGMan. Individual workflow tasks are managed by a workload management framework, HTCondor, which supervises task executions on local and remote resources. Pegasus logs provide time-stamped event instances from these executions. The following example shows the analysis of Pegasus execution logs, stored in a local folder (submit_dir), for a workflow execution using the PegasusLogsParser class:

import pathlib
from wfcommons.wfinstances import PegasusLogsParser

# creating the parser for the Pegasus workflow
submit_dir = pathlib.Path('/path/to/pegasus/submit/dir/seismology/chameleon-100p-001/')
parser = PegasusLogsParser(submit_dir=submit_dir)

# generating the workflow instance object
workflow = parser.build_workflow('pegasus-workflow-test')

# writing the workflow instance to a JSON file
workflow_path = pathlib.Path('./pegasus-workflow.json')
workflow.write_json(workflow_path)

The Instance Analyzer

The InstanceAnalyzer class provides a number of tools for analyzing collection of workflow execution instances. The goal of the InstanceAnalyzer is to perform analyzes of one or multiple workflow execution instances, and build summaries of the analyzes per workflow’ task type prefix.

Warning

Although any workflow execution instance represented as a Instance object (i.e., compatible with WfFormat) can be appended to the InstanceAnalyzer, we strongly recommend that only instances of a single workflow application type be appended to an analyzer object. You may though create several analyzer objects per workflow application.

The append_instance() method allows you to include instances for analysis. The build_summary() method processes all appended instances. The method applies probability distributions fitting to a series of data to find the best (i.e., minimizes the mean square error) probability distribution that represents the analyzed data. The method returns a summary of the analysis of instances in the form of a Python dictionary object in which keys are task prefixes (provided when invoking the method) and values describe the best probability distribution fit for tasks’ runtime, and input and output data file sizes. The code excerpt below shows an example of an analysis summary showing the best fit probability distribution for runtime of the individuals tasks (1000Genome workflow):

"individuals": {
    "runtime": {
        "min": 48.846,
        "max": 192.232,
        "distribution": {
            "name": "skewnorm",
            "params": [
                11115267.652937062,
                -2.9628504044929433e-05,
                56.03957070238482
            ]
        }
    },
    ...
}

Workflow analysis summaries are used by WfChef to develop Workflow Recipes, in which themselves are used to generate realistic synthetic workflow instances.

Probability distribution fits can also be plotted by using the generate_fit_plots() or generate_all_fit_plots() methods – plots will be saved as png files.

Examples

The following example shows the analysis of a set of instances, stored in a local folder, of a Seismology workflow. In this example, we seek for finding the best probability distribution fitting for task prefixes of the Seismology workflow (sG1IterDecon, and wrapper_siftSTFByMisfit), and generate all fit plots (runtime, and input and output files) into the fits folder using seismology as a prefix for each generated plot:

import pathlib
from wfcommons import Instance, InstanceAnalyzer

# obtaining list of instance files in the folder
INSTANCES_PATH = pathlib.Path('/path/to/some/instance/folder/')
instance_files = [f for f in INSTANCES_PATH.glob('*') if INSTANCES_PATH.joinpath(f).is_file()]

# creating the instance analyzer object
analyzer = InstanceAnalyzer()

# appending instance files to the instance analyzer
for instance_file in instance_files:
    instance = Instance(input_instance=INSTANCES_PATH.joinpath(instance_file))
    analyzer.append_instance(instance)

# list of workflow task name prefixes to be analyzed in each instance
workflow_tasks = ['sG1IterDecon', 'wrapper_siftSTFByMisfit']

# building the instance summary
instances_summary = analyzer.build_summary(workflow_tasks, include_raw_data=True)

# generating all fit plots (runtime, and input and output files)
analyzer.generate_all_fit_plots(outfile_prefix='fits/seismology')