Skip to content

Latest commit

 

History

History
71 lines (60 loc) · 3.56 KB

faq.md

File metadata and controls

71 lines (60 loc) · 3.56 KB
Raw

Frequently Asked Questions

Does ODC support multiple active topologies?

Yes. ODC is able to manage multiple DDS sessions concurrently. In this case, each DDS session is an independent active topology. But DDS itself can have only one active topology at a time. Moreover, a number of features in ODC/DDS/FairMQ allow to support a different use case scenarios.

  • DDS topology API allows to create a topology either by editing an XML file or programmatically using C++ or even combining these two approaches. For instance, we have a DDS topology exported from DPL and another topology describing Data Distribution. We can programmatically combine these two topologies into a single topology which can be activated by ODC. Here is an example of how to do it in a couple of lines of code.

  • FairMQ SDK, as well as ODC, allow to address only a part of the topology when changing the states. This is done by providing a regular expression of the task paths which states has to be changed. For instance, you can change only the states of TfBuilder task.

What is a path in DDS topology?

Each task in a DDS topology can be referenced by a path. A topology has a hierarchical structure which is reflected in the path. It is similar to a filesystem path. There are two types of topology paths. First type is used to identify declared topology elements. Second type is used during a runtime. For instance, "main/group1/collection1/task1" is a path for a task1 task declaration. And main/group1/collection1_4/task1_0 is a runtime path of one of the task1 tasks. Note _X. It's an index of a particular task or collection which is assigned to task during the runtime. Each collection and task has a unique path during a runtime.

Having a topology like this:

<topology name="topo1">
    <decltask name="task1">
        <exe>task1.exe</exe>
    </decltask>
    <decltask name="task2">
        <exe>task2.exe</exe>
    </decltask>
    <declcollection name="collection1">
       <tasks>
           <name>task1</name>
           <name n="2">task2</name>
       </tasks>
    </declcollection>
    <main name="main">
       <task>task1</task>
       <collection>collection1</collection>
       <group name="group1" n="2">
          <task>task2</task>
          <collection>collection1</collection>
       </group>
    </main>
</topology>

Results in the following paths of declared tasks:

main/task1
main/collection1/task1
main/collection1/task2
main/group1/task2
main/group1/collection1/task1
main/group1/collection1/task2

For runtime the list of paths is the following:

main/task1_0
main/collection1_0/task1_0
main/collection1_0/task2_0
main/collection1_0/task2_1
main/group1/task2_0
main/group1/collection1_0/task1_0
main/group1/collection1_0/task2_0
main/group1/collection1_0/task2_1
main/group1/task2_1
main/group1/collection1_1/task1_0
main/group1/collection1_1/task2_0
main/group1/collection1_1/task2_1

Empty path refers to a broadcast to all tasks in the topology. If you want to address a particular set of tasks in the topology you can use a regular expression.

Also each task and collection has a hash (uint64_t) which can be used to address a task or a collection.

Paths as well as other runtime information of task and collection is populated as environment variables and available for each task in the topology.

See more on DDS topology here.