What is pScheduler?¶
pScheduler is responsible for managing the execution of network measurements, or more generally tasks, in perfSONAR. When you want to run a network measurement on perfSONAR you ask a pScheduler server, either through the
pscheduler command-line client or through the pScheduler API.
Software that transfers results to long-term storage or elsewhere for further processing.
- Assist Server
A pScheduler server that accepts a test specification and can return the lead participant. The assist server may or may not be involved in the test specified, it’s role is simply to tell you the lead where the task should be submitted. By default the assist server is always the local host.
- Lead Participant
The pScheduler server where a task should be submitted. It will become responsible for contacting other participants, making scheduling decisions based on gathered information and archiving results.
Rules that define who is allowed to run what type of tests
A node that runs pScheduler and handles some facet of conducting a test.
- Schedule Horizon
The amount ahead of time that pScheduler puts a run on the schedule.
A job for pScheduler to do, consisting of a test to be carried out, scheduling information and other options.
A set of the parameters required to perform a unit of work. A unit of work is usually a network measurement in perfSONAR, but could be any type of operation. Tests are generally named in terms of what is being produced, not what does the producing. As an example, if what is being produced is a report of the speed of a data transfer, then the test name is throughput NOT a tool name like iperf.
The pScheduler server will handle the coordination, execution and optionally storage of the task requested. Many of the tools pScheduler executes could be run independently of pScheduler, but the added value provided by pScheduler is significant to perfSONAR hosts for the following reasons:
Measurement Integrity - pScheduler maintains a schedule of all measurements to be run and will not allow any measurements it knowns about to run simultaneously if doing so would adversely affect the result in a significant way. For example, it will not run two throughput tests at the same time as the competition for resources could affect the results of each. In contrast it will run latency tests in the background as the low resource consumption does not significantly affect results of parallel tests.
Simplified Coordination - pScheduler not only simplifies coordination during task execution, but also after you have the result. pScheduler will contact each end and handle bringing up any daemons as required. It also has a plug-in architecture that allows you to send the result elsewhere, such as a long-erm storage system, when the measurement completes.
Access Control - pScheduler has a limits system that allows the definition of rules about who can run what type of measurements.
Diagnostics - pScheduler gives you visibility into the schedule it maintains meaning you can determine when a task ran, runs or will run. It also keeps for some amount of time information about the outcome, including if the result was a failure, which can be useful for diagnosing issues with your test infrastructure.
In addition to these foundational features, perhaps pScheduler greatest asset is it’s extensibility. It allows allows plug-ins for new tests, tools and archivers to be written. This means that pScheduler is not limited by what it can do today, and it’s possible to extend it to perform new type of measurements or other functions as well as to have their results sent to new types of storage and/or analysis tools.
The remainder of this document provides some basic information on pScheduler and pointers to more detailed information where applicable.
Basic Command-Line Client Usage¶
Everything in the pScheduler command-line interface (CLI) begins by running the pscheduler program and specifying a command and, optionally, arguments:
pscheduler COMMAND [ ARGUMENTS ]
At all points along the command line, using the –help switch will cause pScheduler to display a synopsis of the command so far, available options and, where appropriate, examples.
For more information on specific commands and their arguments see:
Running A pScheduler Server¶
If you want tp be able to schedule tasks on your host you will need to run a pScheduler server. For more information on running a pScheduler server and to configure it with limits that meet your needs, see the following documents:
BWCTL Backward Compatibility¶
pScheduler was introduced in perfSONAR version 4.0 and replaced a component called BWCTL. pScheduler is a completely new codebase and protocol. It does NOT speak the BWCTL protocol natively but it does provide a backward compatibility strategy with BWCTL. This strategy does have limitations though and works as follows:
For tests from a host running pScheduler to a host with only BWCTL - For throughput, rtt and trace tests, pScheduler will automatically detect the lack of pScheduler on the remote-end and the presence of BWCTL. It will then choose a pScheduler tool that executes the bwctl, bwping or bwctraceroute command instead of the underlying tool directly. For throughput tests, this has the side effect of requiring more time on the pScheduler schedule since BWCTL may not run the test immediately. If the BWCTL test does not start within 60 seconds of the start time assigned by pScheduler, the test will fail.
For tests from a host running BWCTL to a host with only pScheduler - perfSONAR 4.0 still runs the BWCTL server to support this use case. The pScheduler server knows nothing about the BWCTL server’s schedule and vice versa, so it is possible tests could collide. Depending on how many remote sites initiates tests to you via BWCTL, the risk of this collision may be relatively low.
Starting in 4.1, the BWCTL backward compatibility tools are not included in the default installation. If you would like to install them (along with BWCTL) you can install the perfsonar-bwctl-compat package using your operating system’s package manager.
If you also frequently use BWCTL at the command-line, we suggest you stop and start using the pScheduler command line. For a guide on how to convert your BWCTL commands to pScheduler see BWCTL to pScheduler Command Reference.