Throughput Control, Heartbeat Period and Flow Controllers

[BrunoDatoMeneses]

Hi, I'm having trouble setting flow controlers and the heartbeat period.

I'm designing a multi-agent system and I use fast-dds so that my agents can communicate. Each agent has several publishers and subribers and I would like to controle the overall throughput. I'm using wireshark to monitor the packets (fast dds monitor gives measures only every 3s).

I tested to change the heartbeat period as explained in https://fast-dds.docs.eprosima.com/en/latest/fastdds/use_cases/large_data/large_data.html#tuning-heartbeat-period

Running only one agent, I experienced no change in the througput (see figures below for 200, 3000 and 10000 ms)

With a smaller time update we can see huge peaks and nothing between them.

I also tested with the flow controlers as explained in https://fast-dds.docs.eprosima.com/en/latest/fastdds/use_cases/large_data/large_data.html#flow-controllers

Also running only one agent, I experience no change in the throughput (see figues below for 1000, 500, 100, 50 bps limits)

The higher thoughput before every periodic peaks are the discoveries.

MY CODE:

//ON THE DOMAIN PARTICIPANT

// Limit bps
static const char* flow_controller_name = "custom_flow_controller";
auto custom_flow_control = std::make_sharedeprosima::fastdds::rtps::FlowControllerDescriptor();
custom_flow_control->name = flow_controller_name;
custom_flow_control->scheduler = eprosima::fastdds::rtps::FlowControllerSchedulerPolicy::FIFO;
custom_flow_control->max_bytes_per_period =PUBLISHERS_BPS_LIMIT / 10;
custom_flow_control->period_ms = 100;

DomainParticipantQos participantQos;
participantQos.name("Domain_Participant_EZ_Chains_Agent " + std::to_string(ID_));
participantQos.flow_controllers().push_back(custom_flow_control);

// Increase the buffer size
participantQos.transport().send_socket_buffer_size = 1048576;
participantQos.transport().listen_socket_buffer_size = 4194304;

domainParticipant_ = DomainParticipantFactory::get_instance()->create_participant(0, participantQos);

// ON EVERY PUBLISHER

PublisherQos publisherQos = PUBLISHER_QOS_DEFAULT;
publisher_ = participant_->create_publisher(publisherQos, nullptr);

// Create the DataWriter
PublishModeQosPolicy publish_mode;
publish_mode.kind = ASYNCHRONOUS_PUBLISH_MODE;

DataWriterQos dataWriterQos = DATAWRITER_QOS_DEFAULT;
dataWriterQos.publish_mode(publish_mode);
dataWriterQos.publish_mode().flow_controller_name = "custom_flow_controller";

RTPSReliableWriterQos reliable_writer_qos;
reliable_writer_qos.times.heartbeatPeriod = {HEARTBEAT_PERIOD_SECONDS,HEARTBEAT_PERIOD_NANO_SECONDS};
dataWriterQos.reliable_writer_qos(reliable_writer_qos);

writer_ = publisher_->create_datawriter(topic_, dataWriterQos, &listener_);

My questions are:
Is it possible to lower those peaks by distributing the throughput using the heartbeat period or the flow controlers?
Am I doing something wrong ? Are there some QoS than can be in conflict with these ones ?

Thank you,
Bruno