6.5. Customizing RPC Server request scheduling

When processing an interface inside an IDL file, Fast DDS-Gen generates two method overloads for the creation of a server.

The first overload creates a server with a request scheduling based on a thread pool. Each request will be processed in a separate thread. The number of threads in the pool is specified in the thread_pool_size argument of the method.

The second overload allows the user to inject a custom scheduling strategy for the created server. This is done by creating a custom class implementing the RpcServerSchedulingStrategy interface. A shared pointer to the custom class is then passed in the scheduler argument of the method.

Special care must be taken when implementing a custom scheduling strategy. Calls to schedule_request() will be performed from the thread executing the server’s run() method. This means that incoming messages will not be processed until the execution of schedule_request() finishes. This becomes particularly important for operations that have input feed parameters, since values for input feeds will not be processed while inside that method.

6.5.1. Example

The following code snippet shows two examples of custom RPC server scheduling strategies:

// A scheduling strategy where requests are processed in the same thread where they are received.
// Should not be used in servers with feed operations.
struct DirectRequestScheduling : public eprosima::fastdds::dds::rpc::RpcServerSchedulingStrategy
{
    void schedule_request(
            const std::shared_ptr<eprosima::fastdds::dds::rpc::RpcRequest>& request,
            const std::shared_ptr<eprosima::fastdds::dds::rpc::RpcServer>& server) override
    {
        server->execute_request(request);
    }

    void server_stopped(
            const std::shared_ptr<eprosima::fastdds::dds::rpc::RpcServer>& server) override
    {
        static_cast<void>(server);
    }
};

// A scheduling strategy where each request is processed in a detached thread.
struct DetachedThreadRequestScheduling : public eprosima::fastdds::dds::rpc::RpcServerSchedulingStrategy
{
    void schedule_request(
            const std::shared_ptr<eprosima::fastdds::dds::rpc::RpcRequest>& request,
            const std::shared_ptr<eprosima::fastdds::dds::rpc::RpcServer>& server) override
    {
        std::thread([server, request](){server->execute_request(request);}).detach();
    }

    void server_stopped(
            const std::shared_ptr<eprosima::fastdds::dds::rpc::RpcServer>& server) override
    {
        static_cast<void>(server);
    }
};