A Workflow Management System (WMS) is software that makes all decisions and takes all actions for executing a workflow using cyberinfrastructure services. It is thus a crucial component in every WRENCH simulator. WRENCH does not provide any WMS implementation, but provides the means for developing custom WMSs. This page is meant to provide high-level and detailed information about implementing a WMS in WRENCH. Full API details are provided in the Developer API Reference.
Basic blueprint for a WMS implementation
A WMS implementation needs to use many WRENCH classes, which are accessed by including a single header file:
A WMS implementation must derive the wrench::WMS class, which means that it can override several virtual member functions, but also that a WMS is a service. As such, it has a main() function that goes through a simple loop as follows:
In the next three sections, we give details on how to implement A, B, and C in the code above. To provide context, we make frequent references to the WMS implementations in the example simulators in the examples/ directory. Afterwards are a few sections that highlight features and functionality relevant to WMS development.
A) Obtaining information about services
Discovering running services
The wrench::WMS base class implements a set of member functions named wrench::WMS::getAvailableComputeServices(), wrench::WMS::getAvailableStorageServices(), wrench::WMS::getAvailableNetworkProximityServices(), etc. These member functions return sets of services that can be used by the WMS to execute its workflow. Some of these member functions are templated to retrieve only particular kind of services. For instance, the wrench::WMS::getAvailableComputeServices<T>() takes a template argument to retrieve particular kinds of compute services. In the example simulator in examples/basic-examples/bare-metal-chain, the WMS implementation in OneTaskAtATimeWMS.cpp includes the following call:
This call stores the first of the bare-metal compute services available to the WMS for executing workflow tasks in the compute_service variable. In this example, the simulator always passes exactly one bare-metal service to the WMS, so this code is valid. However, wrench::WMS::getAvailableComputeServices<T>() can return an empty set.
The above member functions (as well as, for instance, wrench::Simulation::add()) return shared pointers (i.e., std::shared_ptr<>) to the service instances. This is to free the developer from the responsibility of freeing memory.
Finding out information about running services
Most service classes provide member functions to get information about the capabilities and properties of the services. For instance, a wrench::ComputeService has a wrench::ComputeService::getNumHosts() member function that returns how many compute hosts the service has access to in total. A wrench::StorageService has a wrench::StorageService::getFreeSpace() member function to find out how many bytes of free space are available on it. And so on...
To take a concrete example, consider the WMS implementation in examples/basic-examples/batch-bag-of-tasks/TwoTasksAtATimeBatchWMS.cpp. This WMS finds out the compute speed of the cores of the compute nodes available to a wrench::BatchComputeService as:
Member function wrench::ComputeService::getCoreFlopRate() returns a map of core compute speeds indexed by hostname (the map thus has one element per compute node available to the service). Since the compute nodes of a batch compute service are homogeneous, the above code simply grabs the core speed value of the first element in the map.
It is important to note that these member functions actually involve communication with the service, and thus incur overhead that is part of the simulation (as if, in the real-world, you would contact a running service with a request for information over the network). This is why the line of code above, in that example WMS, is executed once and the core compute speed is stored in the core_flop_rate variable to be re-used by the WMS repeatedly throughout its execution.
B) Interacting with services
A WMS can have many and complex interactions with services, especially with compute and storage services. In this section, we describe how WRENCH makes these interactions relatively easy, providing examples for each kind of interaction for each kind of service.
Job Manager and Data Movement Manager
As expected, each service type provides its own API. For instance, a network proximity service provides member functions to query the service's host distance databases. The Developer API Reference provides all necessary documentation, which also explains which member functions are synchronous and which are asynchronous (in which case some event will occur in the future). However, the WRENCH developer will find that many member functions that one would expect are nowhere to be found. For instance, the compute services do not have member functions for submitting workflow tasks for execution!
The rationale for the above is that many member functions 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 use data structures to perform the necessary bookkeeping of ongoing service interactions, and process incoming control messages from the services on the (simulated) network or alternately register many callbacks. Instead, WRENCH provides managers. One can think of managers as separate threads that handle all asynchronous interactions with services, and which have been implemented for your convenience to make interacting with services easy.
There are two managers: a job manager (class wrench::JobManager) and a data movement manager (class wrench::DataMovementManager). The base wrench::WMS class provides two member functions for instantiating and starting these managers: wrench::WMS::createJobManager() and wrench::WMS::createDataMovementManager().
Creating these managers typically is the first thing a WMS does. For instance, the WMS in examples/basic-examples/bare-metal-data-movement/DataMovementWMS.cpp starts by doing:
Each manager has its own documented API, and is discussed further in sections below.
Interacting with storage services
The possible interactions between a WMS and a storage service include:
- Synchronously check that a file exists
- Synchronously read a file (rarely used by a WMS but included for completeness)
- Synchronously write a file (rarely used by a WMS but included for completeness)
- Synchronously delete a file
- Synchronously copy a file from one storage service to another
- Asynchronously copy a file from one storage service to another
The first 4 interactions above are done by calling member functions of the wrench::StorageService class. The last two are done via a Data Movement Manager, i.e., by calling member functions of the wrench::DataMovementManager class. Some of these member functions take an optional wrench::FileRegistryService argument, in which case they will also update entries in a file registry service (e.g., removing an entry when a file is deleted).
See this page for concrete examples of interactions with a wrench::SimpleStorageService.
Interacting with compute services
The Job abstraction
The main activity of a WMS is to execute workflow tasks on compute services. Rather than submitting tasks directly to compute services, a WMS must create "jobs", which can comprise multiple tasks and involve data copy/deletion operations. The job abstraction is powerful and greatly simplifies the task of a WMS while affording flexibility.
There are two kinds of jobs in WRENCH: wrench::PilotJob and wrench::StandardJob. A pilot job (sometimes called a "placeholder job" in the literature) 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 it starts it just looks to the WMS like a temporary (bare-metal) compute service to which standard jobs can be submitted.
The most common kind of job is the standard job. A standard job is a unit of execution by which a WMS tells a compute service to do a set of operations. More specifically, in its most complete form, a standard job specifies:
- A set (in fact a vector) of
wrench::WorkflowTaskto execute, so that each task without all its predecessors in the set is ready; - A
std::mapof<wrench::WorkflowFile*, std::shared_ptr<wrench::StorageService>>pairs that specifies from which storage services particular input files should be read and to which storage services output files should be written; - 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 and file copy operations.
Any of the above can actually be empty, and in the extreme a standard job can do nothing.
Standard jobs and pilot jobs are created via the job manager, which provides a wrench::JobManager::createPilotJob() member function and several versions of a wrench::JobManager::createStandardJob() member function. Briefly put, the job manager is a job factory.
The job manager provides the following expected member functions:
wrench::JobManager::submitJob(): asynchronous submission of a job to a compute service.wrench::JobManager::terminateJob(): synchronous termination of a previously submitted job.wrench::JobManager::getPendingPilotJobs(): synchronous retrieval of the list of pending pilot jobs.wrench::JobManager::getRunningPilotJobs(): synchronous retrieval of the list of running pilot jobs.wrench::JobManager::forgetJob(): free memory for a completed/failed job.
The next section gives many examples of interactions with each kind of compute service.
Click on the following links to see detailed descriptions and examples of how jobs are submitted to each compute service type:
- Bare-metal compute service
- Batch compute service
- Cloud compute service
- Virtualized cluster compute service
- HTCondor compute service
Interacting with file registry services
Interaction with a file registry service is straightforward and done by directly calling member functions of the wrench::FileRegistryService class. Note that often file registry service entries are managed automatically, e.g., via calls to wrench::DataMovementManager and wrench::StorageService member functions. So often a WMS does not need to interact with the file registry service.
Adding/removing an entry to a file registry service is done as follows:
The wrench::FileLocation class is a convenient abstraction for a file copy available at some storage service.
Retrieving all entries for a given file is done as follows:
If a network proximity service is running, it is possible to retrieve entries for a file sorted by non-decreasing proximity from some reference host. Returned entries are stored in a (sorted) std::map where the keys are network distances to the reference host. For instance:
See the documentation of wrench::FileRegistryService for more API member functions.
Interacting with network proximity services
Querying a network proximity service is straightforward. For instance, to obtain a measure of the network distance between hosts "Host1" and "Host2", one simply does:
This distance corresponds to half the round-trip-time, in seconds, between the two hosts. If the service is configured to use the Vivaldi coordinate-based system, as in our example above, this distance is actually derived from network coordinates, as computed by the Vivaldi algorithm. In this case, one can actually ask for these coordinates for any given host:
See the documentation of wrench::NetworkProximityService for more API member functions.
C) Workflow execution events
Because the WMS performs asynchronous operations, it needs to wait for and re-act to events. This is done by calling the wrench::WMS::waitForAndProcessNextEvent() member function implemented by the base wrench::WMS class. A call to this member function blocks until some event occurs and then calls a callback member function. The possible event classes all derive from the wrench::WorkflowExecutionEvent class, and a WMS can override the callback member function for each possible event (the default member function does nothing but print some log message). These overridable callback member functions are:
wrench::WMS::processEventStandardJobCompletion(): react to a standard job completionwrench::WMS::processEventStandardJobFailure(): react to a standard job failurewrench::WMS::processEventPilotJobStart(): react to a pilot job beginning executionwrench::WMS::processEventPilotJobExpiration(): react to a pilot job expirationwrench::WMS::processEventFileCopyCompletion(): react to a file copy completionwrench::WMS::processEventFileCopyFailure(): react to a file copy failure
Each member function above takes in an event object as parameter. In the case of failure, the event includes a wrench::FailureCause object, which can be accessed to analyze (or just display) the root cause of the failure.
Consider the WMS in examples/basic-examples/bare-metal-bag-of-tasks/TwoTasksAtATimeWMS.cpp. At each each iteration of its main loop it does:
In this simple example, only one of two events could occur at this point: a standard job completion or a standard job failure. As a result, this WMS overrides the two corresponding member functions as follows:
You may note some difference between the above code and that in examples/basic-examples/bare-metal-bag-of-tasks/TwoTasksAtATimeWMS.cpp. This is for clarity purposes, and especially because we have not yet explained how WRENCH does message logging. See an upcoming section about logging.
While the above callbacks are convenient, sometimes it is desirable to do things more manually. That is, wait for an event and then process it in the code of the main loop of the WMS rather than in a callback member function. This is done by calling the wrench::waitForNextEvent() member function. For instance, the WMS in examples/basic-examples/bare-metal-data-movement/DataMovementWMS.cpp does it as:
Exceptions
Most member functions in the WRENCH Developer API throw exceptions. In fact, most of the code fragments above should be in try-catch clauses, catching these exceptions.
Some exceptions correspond to failures during the simulated workflow executions (i.e., errors that would occur in a real-world execution and are thus part of the simulation). 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 misuses of the WRENCH API or internal WRENCH errors.
Finding information and interacting with hardware resources
The wrench::Simulation class provides many member functions to discover information about the (simulated) hardware platform and interact with it. Some of these member functions are static, but others are not. The wrench:WMS class includes a simulation object. Thus, the WMS can call member functions on the this->simulation object. For instance, this fragment of code shows how a WMS can check that a host exists (given a hostname) and, if so, set its pstate (power state) to the highest possible setting.
See the documentation of the wrench::Simulation class for all details. Specifically regarding host pstates, see the example WMS in examples/basic-examples/cloud-bag-of-tasks-energy/TwoTasksAtATimeCloudWMS.cpp, which interacts with host pstates (and the examples/basic-examples/cloud-bag-of-tasks-energy/four_hosts_energy.xml platform description file which defines pstates).
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):
wrench::StandardJobScheduler: A class that has awrench::StandardJobScheduler::scheduleTasks()member function (to be overwritten) that can be invoked at any time by the WMS to submit tasks (inside standard jobs) to compute services.wrench::PilotJobScheduler: A class that has awrench::PilotJobScheduler::schedulePilotJobs()member function (to be overwritten) that can be invoked at any time by the WMS to submit pilot jobs to compute services.
Although not required, it is possible to implement most (or even all) decision-making in these two member functions so as to have a clean separation of concern between the decision-making part of the WMS and the rest of its functionality. This kind of design is used in the example simulators in the examples/real-workflow-example/ directory.
Logging
It is typically 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 C's 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() member function, which takes as parameter a color specification:
wrench::TerminalOutput::COLOR_BLACKwrench::TerminalOutput::COLOR_REDwrench::TerminalOutput::COLOR_GREENwrench::TerminalOutput::COLOR_YELLOWwrench::TerminalOutput::COLOR_BLUEwrench::TerminalOutput::COLOR_MAGENTAwrench::TerminalOutput::COLOR_CYANwrench::TerminalOutput::COLOR_WHITE
When inspecting the code of the WMSs in the example simulators you will find many examples of calls to wrench::WRENCH_INFO(). The logging is per .cpp file, each of which corresponds to a declared logging category. For instance, in examples/basic-examples/batch-bag-of-tasks/TwoTasksAtATimeBatchWMS.cpp, you will find the typical pattern:
The name of the logging category, in this case custom_wms, can then be passed to the --log command-line argument. For instance, invoking the simulator with additional argument --log=custom_wms.threshold=info will make it so that only those WRENCH_INFO statements in TwoTasksAtATimeBatchWMS.cpp will be printed (in green!).