WRENCH 101
Developer Documentation

WRENCH 101 is a page and a set of documents that provide detailed information for each WRENCH's classes of users, and higher-level content than the API Reference. For instructions on how to install, run a first example, or create a basic WRENCH-based simulator, please refer to their respective sections in the documentation.

This Developer 101 guide describes WRENCH's architectural components necessary to build your own WMS (Workflow Management Systems).


10,000-ft view of a simulated WMS

A Workflow Management System (WMS), i.e., the software that makes all decisions and takes all actions for executing a workflow, is implemented in WRENCH as a simulated process. This process has a main() function that goes through a simple loop as follows:

while ( workflow_execution_hasnt_completed_or_failed ) {
// interact with services
// wait for an event
}

Blueprint for a WMS in WRENCH

A WMS implementation in WRENCH must derive the wrench::WMS class, and typically follows the following steps:

  1. Get references to running services: The wrench::WMS base class implements a set of methods named wrench::WMS::getAvailableComputeServices(), wrench::WMS::getAvailableStorageServices(), etc. These methods return sets of services that can be used by the WMS to execute its workflow.
  2. Acquire information about the services: Some service classes provide methods to get information about the capabilities of the services. For instance, a wrench::ComputeService() has a wrench::ComputeService::getNumHosts() method that makes it possible to find out how many compute hosts the service has access to in total. A wrench::StorageService has a wrench::StorageService::getFreeSpace() method to find out have many bytes of free space are available to it. Note that these methods actually involve communication with the service, and thus incur (simulated) overhead.
  3. Go through a main loop: The heart of the WMS's execution consists in going through a loop until the workflow is executed or has failed to execute. This loop consists of two main steps:
    • Interact with services: This is where the WMS can cause workflow files to be copied between storage services, and where workflow tasks can be submitted to compute services. See the Interacting with services section below for more details.
    • Wait for an event: This is where the WMS is waiting for services to reply with "work done" or "work failed" events. See the Workflow execution events section for more details.

Interacting with services

Each service provides its own API. For instance, a network proximity service provides methods to query the service's host distance databases. The API Reference provides all necessary documentation, which also explains which methods are synchronous and which are asynchronous (in which case some event will likely occur in the future). However, the WRENCH developer will find that many methods that one would expect are nowhere to be found. For instance, the compute services do not have methods for compute job submissions!

The rationale for the above is that many methods need to be asynchronous so that the WMS can use services concurrently. For instance, a WMS could submit a compute job to two distinct compute services asynchronously, and then wait for the service which completes its job first and cancel the job on the other service. Exposing this asynchronicity to the WMS would require that the WRENCH developer uses data structures to perform the necessary bookkeeping of ongoing service interactions, and process incoming control messages from the services on the (simulated) network or register many callbacks. Instead, WRENCH provides managers. One can think of managers are separate threads that handle all asynchronous interactions with services, and which have been implemented for your convenience to make service interactions easy.

For now there are two possible managers: a job manager manager (class wrench::JobManager) and a data movement manager (class wrench::DataMovementManager). The base wrench::WMS class provides two methods for instantiating and starting these managers: wrench::WMS::createJobManager() and wrench::WMS::createDataMovementManager(). Creating these managers typically is the first thing a WMS does. Each manager has its own documented API, and is discussed further in sections below.

Copying workflow data

The WMS may need to explicitly copy files from one storage service to another storage service, e.g., to improve data locality when executing workflow tasks. File copies are accomplished through the data movement manager, which provides two methods:

Both methods take an optional wrench::FileRegistryService argument, in which case they will also update this file registry service with a new entry once the file copy has been completed.

Running workflow tasks

A workflow comprises tasks, and a WMS must pack tasks into jobs to execute them. There are two kinds of jobs in WRENCH: wrench::PilotJob and wrench::StandardJob. A pilot job (sometimes called a "placeholder job") is a concept that is mostly relevant for batch scheduling. In a nutshell, it is a job that allows late binding of tasks to resources. It is submitted to a compute service (provided that service supports pilot jobs), and when a pilot job starts it just looks to the WMS like a short-lived compute service to which standard jobs can be submitted.

The most common kind of jobs is the standard job. A standard job is a unit of execution by which a WMS tells a compute service to do things. More specifically, in its most complete form, a standard job specifies:

  • A vector of wrench::WorkflowTask to execute;
  • A std::map of <wrench::WorkflowFile*, wrench::StorageService *> values which specifies from which storage services particular input files should be read and to which storage services output files should be written. (Note that a compute service can be associated to a "by default" storage service upon instantiation);
  • A set of file copy operations to be performed before executing the tasks;
  • A set of file copy operations to be performed after executing the tasks; and
  • A set of file deletion operations to be performed after executing the tasks.

Any of the above can actually be empty, and in the extreme a standard job does nothing.

Standard jobs and pilot jobs are created via the job manager (see multiple versions of the wrench::JobManager::createStandardJob() and wrench::JobManager::createPilotJob() methods). The job manager thus acts as a job factory, and provides job management methods:

Workflow execution events

Because the WMS, in part via the managers, performs many asynchronous operations, it needs to act as an event handler. This is called by calling the wrench::WMS::waitForAndProcessNextEvent() method implemented by the base wrench::WMS class. A call to this method blocks until some event occurs. The possible event classes all derive the wrench::WorkflowExecutionEvent class. A WMS can override a method to react to each possible event (the default method does nothing but print some log message). At the time this documentation is being written, these overridable methods are:

Each method above takes in an event object as parameter, and each even class offers several methods to inspect the meaning of the event. In the case of failure, the event includes a wrench::FailureCause object, which can be accessed to understand the root cause of the failure.

Exceptions

Most methods in the WRENCH Developer API throw wrench::WorkflowExecutionException instances when exceptions occur. These are exceptions that corresponds to failures during the simulated workflow executions (i.e., errors that would occur in a real-world execution). Each such exception contains a wrench::FailureCause object, which can be accessed to understand the root cause of the execution failure. Other exceptions (e.g., std::invalid_arguments, std::runtime_error) are thrown as well, which are used for detecting mis-uses of the WRENCH API or internal WRENCH errors.

Schedulers for decision-making

A large part of what a WMS does is make decisions. It is often a good idea for decision-making algorithms (often simply called "scheduling algorithms") to be re-usable across multiple WMS implementations, or plug-and-play-able for a single WMS implementation. For this reason, the wrench::WMS constructor takes as parameters two objects (or null pointers if not needed):

Although not required, it is possible to implement most (or even all) decision-making in these two methods so at to have a clean separation of concern between the decision-making part of the WMS and the rest of its functionality. This kind o design is used in the simple example provided in the examples/simple-example directory.

Logging

It is often desirable for the WMS to print log output to the terminal. This is easily accomplished using the wrench::WRENCH_INFO, wrench::WRENCH_DEBUG, and wrench::WRENCH_WARN macros, which are used just like printf. Each of these macros corresponds to a different logging level in SimGrid. See the SimGrid logging documentation for all details.

Furthermore, one can change the color of the log messages with the wrench::TerminalOutput::setThisProcessLoggingColor() method, which takes as parameter a color specification: