Table of Contents
The first step is to install the WRENCH library, following the instructions on the installation page.
Running a First Example
Typing make
in the top-level directory will compile the examples, and make install
will put the examples binaries in the /usr/local/bin
folder (for MacOS and most Linux distributions).
WRENCH provides a simple example in the examples/simple-example
directory, which generates two executables: a cloud-based example wrench-simple-example-cloud
, and a batch-system-based (e.g., SLURM) example wrench-simple-example-batch
. To run the examples, simply use one of the following commands:
Understanding the Simple Example
Both versions of the example (cloud of batch) require two command-line arguments: (1) a SimGrid virtual platform description file; and (2) a WRENCH workflow file.
- SimGrid simulated platform description file: A SimGrid simulation must be provided with the description of the platform on which an application execution is to be simulated. This is done via a platform description file, in XML, that includes definitions of compute hosts, clusters of hosts, storage resources, network links, routes between hosts, etc. A detailed description on how to create a platform description file can be found here.
- WRENCH workflow file: WRENCH provides native parsers for DAX (DAG in XML) and JSON workflow description file formats. Refer to their respective Web sites for detailed documentation.
The source file for the cloud-based simulator is at examples/simple-example/SimulatorCloud.cpp
and at examples/simple-example/SimulatorBatch.cpp
for the batch-based example. These source files, which are heavily commented, and perform the following:
- The first step is to read and parse the workflow and the platform files, and to create a simulation object (
wrench::Simulation
). - A storage service (
wrench::SimpleStorageService
) is created and deployed on a host. - A cloud (
wrench::CloudService
) or a batch (wrench::BatchService
) service is created and deployed on a host. Both services are seen by the simulation as compute services (wrench::ComputeService
) – jobs can then be submitted to these services. - A Workflow Management System (
wrench::WMS
) is instantiated (in this case theSimpleWMS
) with a reference to a workflow object (wrench::Workflow
) and a scheduler (wrench::Scheduler
). The scheduler implements the decision-making algorithms inside the WMS. These algorithms are modularized (so that the same WMS implementation can be iniated with various decision-making algorithms in different simulations). The source codes for the schedulers, which is of interest to "Developers" (i.e., those users who use the WRENCH Developer API), is in directoryexamples/scheduler
. - A file registry (
wrench::FileRegistryService
), a.k.a. a file replica catalog, which keeps track of files stored in different storage services, is deployed on a host. - Workflow input files are staged on the storage service
- The simulation is launched, executes, and completes.
- Timestamps can be retrieved to analyze the simulated execution.
This simple example can be used as a blueprint for starting a large WRENCH-based simulation project. The next section provides further details about this process.
Preparing the Environment
Internal developers are expected to contribute code to WRENCH's core components. Please, refer to the API Reference to find the detailed documentation for WRENCH functions.
Note: It is strongly recommended that WRENCH internal developers (contributors) fork WRENCH's code from the GitHub repository, and create pull requests with their proposed modifications.
WRENCH Directory and File Structure
WRENCH follows a standard C++ project directory and files structure: