Skip to content

Latest commit

 

History

History
413 lines (334 loc) · 17.5 KB

README.md

File metadata and controls

413 lines (334 loc) · 17.5 KB

iceperf

Introduction

!!! note Since not all IPC mechanisms are supported on all platforms the IPC benchmark only runs fully on QNX and Linux. The iceoryx C or C++ API related benchmark is supported on all platforms.

This example measures the latency of IPC transmissions between two applications. We compare the latency of iceoryx with message queues and unix domain sockets.

The measurement is carried out with several payload sizes. Round trips are performed for each payload size, using either the default setting or the provided command line parameter for the number of round trips to do. The time measurement only considers the time to allocate/release memory and the time to send the data. The construction and initialization of the payload is not part of the measurement.

At the end of the benchmark, the average latency for each payload size is printed.

Run iceperf

Create three terminals and run one command in each of them. In this setup the leader is doing the ping pong measurements with the follower. You can set the number of measurement iterations (number of round trips) with a command line parameter of iceperf-bench-leader (e.g. ./iceperf-bench-leader -n 100000). There are further options which can be printed by calling ./iceperf-bench-leader -h.

    # If installed and available in PATH environment variable
    iox-roudi
    # If build from scratch with script in tools
    $ICEORYX_ROOT/build/install/prefix/bin/iox-roudi

    build/iceoryx_examples/iceperf/iceperf-bench-follower

    build/iceoryx_examples/iceperf/iceperf-bench-leader

If you would like to test only the C++ API or the C API you can start iceperf-bench-leader with the parameter -t iceoryx-cpp-api or -t iceoryx-c-api.

    build/iceoryx_examples/iceperf/iceperf-bench-follower

    build/iceoryx_examples/iceperf/iceperf-bench-leader -n 100000 -t iceoryx-cpp-api

Expected Output

The measured transmission modes depend on the operating system (e.g. no message queue on MacOS). The measurements depend on the benchmark parameters and the hardware.

The following shows an example output with Ubuntu 18.04 on Intel(R) Xeon(R) CPU E3-1505M v5 @ 2.80GHz.

iceperf-bench-leader Application

******   MESSAGE QUEUE    ********
Waiting for: subscription, subscriber [ success ]
Measurement for: 1 kB, 2 kB, 4 kB, 8 kB, 16 kB, 32 kB, 64 kB, 128 kB, 256 kB,
512 kB, 1024 kB, 2048 kB, 4096 kB
Waiting for: unsubscribe  [ finished ]

#### Measurement Result ####
100000 round trips for each payload.

| Payload Size [kB] | Average Latency [µs] |
|------------------:|---------------------:|
|                 1 |                  3.1 |
|                 2 |                  3.2 |
|                 4 |                  3.8 |
|                 8 |                  5.2 |
|                16 |                  7.7 |
|                32 |                   13 |
|                64 |                   23 |
|               128 |                   43 |
|               256 |                   81 |
|               512 |              1.6e+02 |
|              1024 |                3e+02 |
|              2048 |              5.9e+02 |
|              4096 |              1.2e+03 |

Finished!

****** UNIX DOMAIN SOCKET ********
Waiting for: subscription, subscriber [ success ]
Measurement for: 1 kB, 2 kB, 4 kB, 8 kB, 16 kB, 32 kB, 64 kB, 128 kB, 256 kB,
512 kB, 1024 kB, 2048 kB, 4096 kB
Waiting for: unsubscribe  [ finished ]

#### Measurement Result ####
100000 round trips for each payload.

| Payload Size [kB] | Average Latency [µs] |
|------------------:|---------------------:|
|                 1 |                  4.3 |
|                 2 |                  4.3 |
|                 4 |                  4.6 |
|                 8 |                    6 |
|                16 |                  8.7 |
|                32 |                   14 |
|                64 |                   27 |
|               128 |                   53 |
|               256 |              1.1e+02 |
|               512 |              2.1e+02 |
|              1024 |              4.2e+02 |
|              2048 |              8.4e+02 |
|              4096 |              1.7e+03 |

Finished!

******      ICEORYX       ********
Waiting for: subscription, subscriber [ success ]
Measurement for: 1 kB, 2 kB, 4 kB, 8 kB, 16 kB, 32 kB, 64 kB, 128 kB, 256 kB,
512 kB, 1024 kB, 2048 kB, 4096 kB
Waiting for: unsubscribe  [ finished ]

#### Measurement Result ####
100000 round trips for each payload.

| Payload Size [kB] | Average Latency [µs] |
|------------------:|---------------------:|
|                 1 |                 0.73 |
|                 2 |                 0.58 |
|                 4 |                 0.61 |
|                 8 |                 0.61 |
|                16 |                 0.59 |
|                32 |                 0.62 |
|                64 |                  0.6 |
|               128 |                 0.58 |
|               256 |                 0.61 |
|               512 |                 0.61 |
|              1024 |                 0.58 |
|              2048 |                 0.61 |
|              4096 |                 0.61 |

Finished!

******   ICEORYX C API    ********
Waiting for: subscription, subscriber [ success ]
Measurement for: 1 kB, 2 kB, 4 kB, 8 kB, 16 kB, 32 kB, 64 kB, 128 kB, 256 kB,
512 kB, 1024 kB, 2048 kB, 4096 kB
Waiting for: unsubscribe  [ finished ]

#### Measurement Result ####
100000 round trips for each payload.

| Payload Size [kB] | Average Latency [µs] |
|------------------:|---------------------:|
|                 1 |                 0.73 |
|                 2 |                 0.58 |
|                 4 |                 0.61 |
|                 8 |                 0.61 |
|                16 |                 0.59 |
|                32 |                 0.62 |
|                64 |                  0.6 |
|               128 |                 0.58 |
|               256 |                 0.61 |
|               512 |                 0.61 |
|              1024 |                 0.58 |
|              2048 |                 0.61 |
|              4096 |                 0.61 |

Finished!

iceperf-bench-follower Application

******   MESSAGE QUEUE    ********
registering with the leader

****** UNIX DOMAIN SOCKET ********
registering with the leader

******      ICEORYX       ********
Waiting for: subscription, subscriber [ success ]
Waiting for: unsubscribe  [ finished ]

******   ICEORYX C API    ********
Waiting for: subscription, subscriber [ success ]
Waiting for: unsubscribe  [ finished ]

Code Walkthrough

Here we briefly describe the setup for performing the measurements in iceperf_bench_leader.hpp/cpp and iceperf_bench_follower.hpp/cpp. Things like initialization, sending and receiving of data are technology specific and can be found in the respective files (e.g. uds.cpp for unix domain socket). Our focus here is on the top-most abstraction layer which allows us to add new IPC technologies to extend and compare them.

iceperf-bench-leader Application Code

Apart from headers for the different IPC technologies, the topic_data.hpp file is included which contains the PerSettings and PerTopic structs. These are used to transfer some information between the applications. The PerTopic struct is used as some kind of header in each transferred sample and is independent of the payload size.

struct PerfSettings
{
    Benchmark benchmark{Benchmark::ALL};
    Technology technology{Technology::ALL};
    uint64_t numberOfSamples{10000U};
};

struct PerfTopic
{
    uint32_t payloadSize{0};
    uint32_t subPackets{0};
    RunFlag runFlag{RunFlag::RUN};
};

The PerfSettings struct is used to synchronize the settings between the leader and the follower application.

The PerfTopic struct is used to share some information during the measurement. It contains payloadSize to specify the payload size used for the current measurement. If it is not possible to transmit the payloadSize with a single data transfer (e.g. OS limit for the payload of a single socket send), the payload is divided into several sub-packets. This is indicated with subPackets. The runFlag is used to shut down the iceperf-bench follower at the end of the benchmark.

Let's use some constants to prevent magic values and set and names for the communication resources that are used.

constexpr const char APP_NAME[]{"iceperf-bench-leader"};
constexpr const char PUBLISHER[]{"Leader"};
constexpr const char SUBSCRIBER[]{"Follower"};

The IcePerfLeader c'tor does a cleanup of potentially outdated resources of technologies which might have left some resources in the file system after an abnormal terminations.

#ifndef __APPLE__
MQ::cleanupOutdatedResources(PUBLISHER, SUBSCRIBER);
#endif
UDS::cleanupOutdatedResources(PUBLISHER, SUBSCRIBER);

The doMeasurement() method executes a measurement for the provided IPC technology and number of round trips. To be able to always perform the same steps and avoiding code duplications, we use a base class with technology independent functionality and the technology has to implement the technology dependent part.

void IcePerfLeader::doMeasurement(IcePerfBase& ipcTechnology) noexcept
{
    ipcTechnology.initLeader();

    std::vector<std::tuple<uint32_t, iox::units::Duration>> latencyMeasurements;
    const std::vector<uint32_t> payloadSizesInKB{1, 2, 4, 8, 16, 32, 64, 128, 256, 512, 1024, 2048, 4096};
    std::cout << "Measurement for:";
    const char* separator = " ";
    for (const auto payloadSizeInKB : payloadSizesInKB)
    {
        std::cout << separator << payloadSizeInKB << " kB" << std::flush;
        separator = ", ";
        auto payloadSizeInBytes = payloadSizeInKB * IcePerfBase::ONE_KILOBYTE;

        ipcTechnology.preLatencyPerfTestLeader(payloadSizeInBytes);

        auto latency = ipcTechnology.latencyPerfTestLeader(m_settings.numberOfSamples);

        latencyMeasurements.push_back(std::make_tuple(payloadSizeInKB, latency));

        ipcTechnology.postLatencyPerfTestLeader();
    }
    std::cout << std::endl;

    ipcTechnology.releaseFollower();

    ipcTechnology.shutdown();

    std::cout << std::endl;
    std::cout << "#### Measurement Result ####" << std::endl;
    std::cout << m_settings.numberOfSamples << " round trips for each payload." << std::endl;
    std::cout << std::endl;
    std::cout << "| Payload Size [kB] | Average Latency [µs] |" << std::endl;
    std::cout << "|------------------:|---------------------:|" << std::endl;
    for (const auto& latencyMeasuement : latencyMeasurements)
    {
        auto payloadSizeInKB = std::get<0>(latencyMeasuement);
        auto latencyInMicroseconds = static_cast<double>(std::get<1>(latencyMeasuement).toNanoseconds()) / 1000.0;
        std::cout << "| " << std::setw(17) << payloadSizeInKB << " | " << std::setw(20) << std::setprecision(2)
                  << latencyInMicroseconds << " |" << std::endl;
    }

    std::cout << std::endl;
    std::cout << "Finished!" << std::endl;
}

Initialization is different for each IPC technology. Here we have to create sockets, message queues or iceoryx publisher and subscriber. With ipcTechnology.initLeader() we set up these resources on the leader side. After the definition of the different payload sizes to use, we execute a single round trip measurement for each individual payload size. The leader has to orchestrate the whole process and has a pre- and post-step for each round trip measurement. ipcTechnology.preLatencyPerfTestLeader(...) sets the payload size for the upcoming measurement. ipcTechnology.latencyPerfTestLeader(m_settings.numberOfSamples) performs the data exchange between leader and follower and returns the time it took to transmit the number of samples in a round trip. After the measurements are taken for each payload size, ipcTechnology.releaseFollower() releases the follower. This is required since the follower is not aware of the benchmark settings, e.g. how many payload sizes are considered and hence we need to issue a shutdown. We clean up the communication resources with ipcTechnology.shutdown() before we print the results.

In the run() method we create instances for the different IPC technologies we want to compare. Each technology is implemented in its own class and implements the pure virtual functions provided with the IcePerfBase class. Before this is done, we send the PerfSettings to the follower application.

int IcePerfLeader::run() noexcept
{
    iox::runtime::PoshRuntime::initRuntime(APP_NAME);

    iox::capro::ServiceDescription serviceDescription{"IcePerf", "Settings", "Generic"};
    iox::popo::PublisherOptions options;
    options.historyCapacity = 1U;
    iox::popo::Publisher<PerfSettings> settingsPublisher{serviceDescription, options};
    if (!settingsPublisher.publishCopyOf(m_settings))
    {
        std::cerr << "Could not send settings to follower!" << std::endl;
        return EXIT_FAILURE;
    }
    // ...
    return EXIT_SUCCESS;
}

Now we can create an object for each IPC technology that we want to evaluate and call the doMeasurement() method.

int IcePerfLeader::run() noexcept
{
    iox::runtime::PoshRuntime::initRuntime(APP_NAME);
    // ...
    if (m_settings.technology == Technology::ALL || m_settings.technology == Technology::POSIX_MESSAGE_QUEUE)
    {
#ifndef __APPLE__
        std::cout << std::endl << "******   MESSAGE QUEUE    ********" << std::endl;
        MQ mq(PUBLISHER, SUBSCRIBER);
        doMeasurement(mq);
#else
        if (m_settings.technology == Technology::POSIX_MESSAGE_QUEUE)
        {
            std::cout << "The message queue is not supported on macOS and will be skipped!" << std::endl;
        }
#endif
    }

    if (m_settings.technology == Technology::ALL || m_settings.technology == Technology::UNIX_DOMAIN_SOCKET)
    {
        std::cout << std::endl << "****** UNIX DOMAIN SOCKET ********" << std::endl;
        UDS uds(PUBLISHER, SUBSCRIBER);
        doMeasurement(uds);
    }

    if (m_settings.technology == Technology::ALL || m_settings.technology == Technology::ICEORYX_CPP_API)
    {
        std::cout << std::endl << "******      ICEORYX       ********" << std::endl;
        Iceoryx iceoryx(PUBLISHER, SUBSCRIBER);
        doMeasurement(iceoryx);
    }

    if (m_settings.technology == Technology::ALL || m_settings.technology == Technology::ICEORYX_C_API)
    {
        std::cout << std::endl << "******   ICEORYX C API    ********" << std::endl;
        IceoryxC iceoryxc(PUBLISHER, SUBSCRIBER);
        doMeasurement(iceoryxc);
    }

    return EXIT_SUCCESS;
}

iceperf_bench_follower Application

The iceperf-bench-follower application is similar to iceperf-bench-leader. The first change is that the SUBSCRIBER and PUBLISHER switch their names.

constexpr const char APP_NAME[]{"iceperf-bench-follower"};
constexpr const char PUBLISHER[]{"Follower"};
constexpr const char SUBSCRIBER[]{"Leader"};

While the run() method of the leader publishes the PerfSettings, the follower is subscribed to those settings and waits for them before the technologies are created, which is done similarly as for the leader.

int IcePerfFollower::run() noexcept
{
    iox::runtime::PoshRuntime::initRuntime(APP_NAME);

    iox::capro::ServiceDescription serviceDescription{"IcePerf", "Settings", "Generic"};
    iox::popo::SubscriberOptions options;
    options.historyRequest = 1U;
    iox::popo::Subscriber<PerfSettings> settingsSubscriber{serviceDescription, options};

    m_settings = getSettings(settingsSubscriber);
    // ...
    return EXIT_SUCCESS;
}

The doMeasurement() method is much simpler than the one from the leader, since it only has to react on incoming data. Apart from ipcTechnology.initFollower() and ipcTechnology.shutdown() all the functionality to perform the round trip for different payload sizes is contained in ipcTechnology.latencyPerfTestFollower()

void IcePerfFollower::doMeasurement(IcePerfBase& ipcTechnology) noexcept
{
    ipcTechnology.initFollower();

    ipcTechnology.latencyPerfTestFollower();

    ipcTechnology.shutdown();
}
[Check out iceperf on GitHub :fontawesome-brands-github:](https://github.com/eclipse-iceoryx/iceoryx/tree/master/iceoryx_examples/iceperf){ .md-button }