使用gRPC基于Protobuf传输大文件或数据流

程序员小涛

发布于 2024-05-26 17:32:21

2970

发布于 2024-05-26 17:32:21

文章被收录于专栏：涛的程序人生涛的程序人生

使用gRPC基于Protobuf传输大文件或数据流

在现代软件开发中，性能通常是关键的考虑因素之一，尤其是在进行大文件传输时。高效的协议和工具可以显著提升传输速度和可靠性。本文详细介绍如何使用gRPC和Protobuf进行大文件传输，并与传统TCP传输进行性能比较。

1. 背景和技术选择

在过去，大文件传输常常使用传统的TCP/IP协议，虽然简单但在处理大规模数据传输时往往速度较慢，尤其在网络条件不佳的环境下更是如此。在最近一个项目中，就有传输大数据文件的需求，用传统方式进行测试发现传输延时无法达到要求，于是在网上查阅资料发现了一个较优的解决方案。

相对于TCP，gRPC提供了一种现代的、高性能的解决方案。gRPC是一个高性能的远程过程调用（RPC）框架，由Google主导开发，使用HTTP/2作为传输层协议，支持多种开发语言，如C++, Java, Python和Go等。Protobuf（Protocol Buffers）则是一种轻量级的数据交换格式，可以高效地序列化结构化数据。

1.1 gRPC的优势

高性能: 利用HTTP/2协议，支持多路复用、服务器推送等现代网络技术。
跨语言支持: 支持多种编程语言，便于在不同的系统间交互。
接口定义: 使用.proto文件定义服务，自动生成服务端和客户端代码，减少重复工作量。
流控制: 支持流式传输数据，适合大文件传输和实时数据处理。

1.2 Protocol Buffers的优势

高效: 编码和解码迅速，且生成的数据包比XML小3到10倍。
灵活: 支持向后兼容性，新旧数据格式可以无缝转换。
简洁: 简化了复杂数据结构的处理，易于开发者使用。

2. 项目配置与环境搭建

为了使用gRPC进行项目开发，首先需要在开发环境中安装gRPC及其依赖的库。以下是gRPC安装的步骤，适用于多种操作系统，包括Windows、Linux和macOS。

2.1 安装gRPC和Protocol Buffers

gRPC的安装可以通过多种方式进行，包括使用包管理器或从源代码编译。以下介绍Ubuntu下安装C++版本的gRPC（捆绑了Protocol Buffers）

注：如果gRPC和Protocol Buffers的版本不匹配会有问题，无法正常使用

2.1.1 安装Cmake

sudo apt install -y cmake

2.1.2 设置环境变量

export MY_INSTALL_DIR=$HOME/.local
mkdir -p $MY_INSTALL_DIR
export PATH="$MY_INSTALL_DIR/bin:$PATH" # 为了永久生效可以将该命令写入~/.bashrc文件中

2.1.3 安装必要的依赖

sudo apt install -y build-essential autoconf libtool pkg-config

2.1.4 下载gRPC源码

git clone --recurse-submodules -b v1.62.0 --depth 1 --shallow-submodules https://github.com/grpc/grpc

2.1.5 编译gRPC和 Protocol Buffers

cd grpc
mkdir -p cmake/build
pushd cmake/build
cmake -DgRPC_INSTALL=ON \
      -DgRPC_BUILD_TESTS=OFF \
      -DCMAKE_INSTALL_PREFIX=$MY_INSTALL_DIR \
      ../..
make -j 4
make install
popd

2.2 CMake配置详解

2.1.1 通用配置

common.cmake 是一个辅助性的 CMake 模块文件，通常用于存放项目中共用的 CMake 配置，以简化和集中管理 CMakeLists.txt 文件中的代码。这种做法有助于提升项目的可维护性和可读性。

在 gRPC 项目中，示例代码中的common.cmake 包括以下内容：

变量设置：定义项目中使用的常见路径和变量，例如 gRPC 和 protobuf 的安装路径，以便在整个项目中重用。
库查找：使用 find_package() 或 find_library() 命令来查找和配置项目所需的依赖库，如 gRPC、protobuf、SSL 等。
编译器选项：统一设置编译器标志，例如 C++ 版本标准、优化级别、警告处理等。
宏定义：创建复用的 CMake 宏或函数，例如用于处理 proto 文件生成相关命令的宏，这有助于避免在 CMakeLists.txt 文件中重复相同的代码块。

# Copyright 2018 gRPC authors.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
# cmake build file for C++ route_guide example.
# Assumes protobuf and gRPC have been installed using cmake.
# See cmake_externalproject/CMakeLists.txt for all-in-one cmake build
# that automatically builds all the dependencies before building route_guide.

cmake_minimum_required(VERSION 3.8)

if(MSVC)
  add_definitions(-D_WIN32_WINNT=0x600)
endif()

find_package(Threads REQUIRED)

if(GRPC_AS_SUBMODULE)
  # One way to build a projects that uses gRPC is to just include the
  # entire gRPC project tree via "add_subdirectory".
  # This approach is very simple to use, but the are some potential
  # disadvantages:
  # * it includes gRPC's CMakeLists.txt directly into your build script
  #   without and that can make gRPC's internal setting interfere with your
  #   own build.
  # * depending on what's installed on your system, the contents of submodules
  #   in gRPC's third_party/* might need to be available (and there might be
  #   additional prerequisites required to build them). Consider using
  #   the gRPC_*_PROVIDER options to fine-tune the expected behavior.
  #
  # A more robust approach to add dependency on gRPC is using
  # cmake's ExternalProject_Add (see cmake_externalproject/CMakeLists.txt).

  # Include the gRPC's cmake build (normally grpc source code would live
  # in a git submodule called "third_party/grpc", but this example lives in
  # the same repository as gRPC sources, so we just look a few directories up)
  add_subdirectory(../../.. ${CMAKE_CURRENT_BINARY_DIR}/grpc EXCLUDE_FROM_ALL)
  message(STATUS "Using gRPC via add_subdirectory.")

  # After using add_subdirectory, we can now use the grpc targets directly from
  # this build.
  set(_PROTOBUF_LIBPROTOBUF libprotobuf)
  set(_REFLECTION grpc++_reflection)
  set(_ORCA_SERVICE grpcpp_orca_service)
  if(CMAKE_CROSSCOMPILING)
    find_program(_PROTOBUF_PROTOC protoc)
  else()
    set(_PROTOBUF_PROTOC $<TARGET_FILE:protobuf::protoc>)
  endif()
  set(_GRPC_GRPCPP grpc++)
  if(CMAKE_CROSSCOMPILING)
    find_program(_GRPC_CPP_PLUGIN_EXECUTABLE grpc_cpp_plugin)
  else()
    set(_GRPC_CPP_PLUGIN_EXECUTABLE $<TARGET_FILE:grpc_cpp_plugin>)
  endif()
elseif(GRPC_FETCHCONTENT)
  # Another way is to use CMake's FetchContent module to clone gRPC at
  # configure time. This makes gRPC's source code available to your project,
  # similar to a git submodule.
  message(STATUS "Using gRPC via add_subdirectory (FetchContent).")
  include(FetchContent)
  FetchContent_Declare(
    grpc
    GIT_REPOSITORY https://github.com/grpc/grpc.git
    # when using gRPC, you will actually set this to an existing tag, such as
    # v1.25.0, v1.26.0 etc..
    # For the purpose of testing, we override the tag used to the commit
    # that's currently under test.
    GIT_TAG        vGRPC_TAG_VERSION_OF_YOUR_CHOICE)
  FetchContent_MakeAvailable(grpc)

  # Since FetchContent uses add_subdirectory under the hood, we can use
  # the grpc targets directly from this build.
  set(_PROTOBUF_LIBPROTOBUF libprotobuf)
  set(_REFLECTION grpc++_reflection)
  set(_PROTOBUF_PROTOC $<TARGET_FILE:protoc>)
  set(_GRPC_GRPCPP grpc++)
  if(CMAKE_CROSSCOMPILING)
    find_program(_GRPC_CPP_PLUGIN_EXECUTABLE grpc_cpp_plugin)
  else()
    set(_GRPC_CPP_PLUGIN_EXECUTABLE $<TARGET_FILE:grpc_cpp_plugin>)
  endif()
else()
  # This branch assumes that gRPC and all its dependencies are already installed
  # on this system, so they can be located by find_package().

  # Find Protobuf installation
  # Looks for protobuf-config.cmake file installed by Protobuf's cmake installation.
  option(protobuf_MODULE_COMPATIBLE TRUE)
  find_package(Protobuf CONFIG REQUIRED)
  message(STATUS "Using protobuf ${Protobuf_VERSION}")

  set(_PROTOBUF_LIBPROTOBUF protobuf::libprotobuf)
  set(_REFLECTION gRPC::grpc++_reflection)
  if(CMAKE_CROSSCOMPILING)
    find_program(_PROTOBUF_PROTOC protoc)
  else()
    set(_PROTOBUF_PROTOC $<TARGET_FILE:protobuf::protoc>)
  endif()

  # Find gRPC installation
  # Looks for gRPCConfig.cmake file installed by gRPC's cmake installation.
  find_package(gRPC CONFIG REQUIRED)
  message(STATUS "Using gRPC ${gRPC_VERSION}")

  set(_GRPC_GRPCPP gRPC::grpc++)
  if(CMAKE_CROSSCOMPILING)
    find_program(_GRPC_CPP_PLUGIN_EXECUTABLE grpc_cpp_plugin)
  else()
    set(_GRPC_CPP_PLUGIN_EXECUTABLE $<TARGET_FILE:gRPC::grpc_cpp_plugin>)
  endif()
endif()

2.2.2 项目配置

这个配置文件包括了从proto文件生成C++代码的命令，以及编译这些生成的源代码文件为库和可执行文件的命令。利用CMake，我们能够确保项目在不同环境中具有可重复构建的能力。

cmake_minimum_required(VERSION 3.15)

project(grpcDemo)

set(CMAKE_CXX_STANDARD 17)

include(common.cmake)

# Proto file
get_filename_component(transfer_proto "transferfile.proto" ABSOLUTE)
get_filename_component(transfer_proto_path "${transfer_proto}" PATH)

# Generated sources
set(transfer_proto_srcs "${CMAKE_CURRENT_BINARY_DIR}/transferfile.pb.cc")
set(transfer_proto_hdrs "${CMAKE_CURRENT_BINARY_DIR}/transferfile.pb.h")
set(transfer_grpc_srcs "${CMAKE_CURRENT_BINARY_DIR}/transferfile.grpc.pb.cc")
set(transfer_grpc_hdrs "${CMAKE_CURRENT_BINARY_DIR}/transferfile.grpc.pb.h")
add_custom_command(
      OUTPUT "${transfer_proto_srcs}" "${transfer_proto_hdrs}" "${transfer_grpc_srcs}" "${transfer_grpc_hdrs}"
      COMMAND ${_PROTOBUF_PROTOC}
      ARGS --grpc_out "${CMAKE_CURRENT_BINARY_DIR}"
        --cpp_out "${CMAKE_CURRENT_BINARY_DIR}"
        -I "${transfer_proto_path}"
        --plugin=protoc-gen-grpc="${_GRPC_CPP_PLUGIN_EXECUTABLE}"
        "${transfer_proto}"
      DEPENDS "${transfer_proto}")

# Include generated *.pb.h files
include_directories("${CMAKE_CURRENT_BINARY_DIR}")

# transfer_grpc_proto
add_library(transfer_grpc_proto
  ${transfer_grpc_srcs}
  ${transfer_grpc_hdrs}
  ${transfer_proto_srcs}
  ${transfer_proto_hdrs})
target_link_libraries(transfer_grpc_proto
  ${_REFLECTION}
  ${_GRPC_GRPCPP}
  ${_PROTOBUF_LIBPROTOBUF})


# Targets greeter_[async_](client|server)
foreach(_target
  transfer_client transfer_server)
  add_executable(${_target} "${_target}.cpp")
  target_link_libraries(${_target}
    transfer_grpc_proto
    ${_REFLECTION}
    ${_GRPC_GRPCPP}
    ${_PROTOBUF_LIBPROTOBUF})
endforeach()

2.3 服务定义与proto文件

在gRPC中，服务和消息的定义是通过.proto文件进行的。例如，定义一个文件传输服务，可以在transferfile.proto中如下定义：

syntax = "proto3";

package filetransfer;

service FileTransferService {
  rpc Upload(stream FileChunk) returns (UploadStatus) {}
}

message FileChunk {
  bytes content = 1;
}

message UploadStatus {
  bool success = 1;
  string message = 2;
}

这里定义了一个FileTransferService服务，包含了一个Upload方法，该方法接受一个FileChunk类型的流，并返回一个UploadStatus状态。

3. 客户端和服务端的实现

客户端和服务端的实现是通过gRPC框架生成的接口进行的，这些接口基于前面定义的.proto文件。

3.1 gRPC客户端实现

客户端的主要职责是打开文件，读取数据，然后以流的形式发送到服务端。实现代码如下：

#include <iostream>
#include <string>
#include <fstream>
#include <chrono>
#include <grpcpp/grpcpp.h>
#include <grpcpp/channel.h>
#include <grpcpp/client_context.h>
#include <grpcpp/create_channel.h>
#include <grpcpp/security/credentials.h>
#include "transferfile.grpc.pb.h"

using grpc::Channel;
using grpc::ClientContext;
using grpc::ClientWriter;
using grpc::Status;
using transferfile::FileChunk;
using transferfile::FileUploadStatus;
using transferfile::TransferFile;

#define CHUNK_SIZE 3 * 1024 * 1024 // 3MB

class TransferClient
{
public:
    TransferClient(std::shared_ptr<Channel> channel) : stub_(TransferFile::NewStub(channel)) {}

    void uploadFile(const std::string &filename);

private:
    std::unique_ptr<TransferFile::Stub> stub_;
};

void TransferClient::uploadFile(const std::string &filename)
{
    FileChunk chunk;
    char *buffer = new char[CHUNK_SIZE];
    FileUploadStatus status;
    ClientContext context;
    std::ifstream infile;
    unsigned long len = 0;

    auto start = std::chrono::steady_clock::now();
    infile.open(filename, std::ios::binary | std::ios::in);
    if (!infile.is_open())
    {
        std::cerr << "Error: File not found" << std::endl;
        return;
    }

    std::unique_ptr<ClientWriter<FileChunk>> writer(stub_->UploadFile(&context, &status)); // Create a writer to send chunks of file
    while (!infile.eof())
    {
        infile.read(buffer, CHUNK_SIZE);
        chunk.set_buffer(buffer, infile.gcount());
        if (!writer->Write(chunk))
        {
            std::cerr << "Error: Failed to write chunk" << std::endl;
            break;
        }
        len += infile.gcount();
    }

    infile.close();
    delete[] buffer;

    writer->WritesDone();
    Status status1 = writer->Finish();
    if (status1.ok() && len != status.length())
    {
        auto end = std::chrono::steady_clock::now();
        std::cout << "File uploaded successfully" << std::endl;
        std::cout << "Time taken: " << std::chrono::duration_cast<std::chrono::milliseconds>(end - start).count() << "ms" << std::endl;
        std::cout << "File size: " << len << " bytes" << std::endl;
        auto speed = (len / std::chrono::duration_cast<std::chrono::milliseconds>(end - start).count()) * 1000 / 1024 / 1024;
        std::cout << "Speed: " << speed << " MB/s" << std::endl;
    }
    else
    {
        std::cerr << "Error: " << status1.error_message() << "len: " << len << " status.length(): " << status.length()
                  << std::endl;
    }
}

int main(int argc, char *argv[])
{
    if (argc < 2)
    {
        std::cerr << "Usage: " << argv[0] << " <filename> [server_ip]" << std::endl;
        return 1;
    }

    std::string filename(argv[1]);
    std::string server_ip = "localhost";

    if (argc == 3)
    {
        server_ip = argv[2];
    }

    TransferClient client(grpc::CreateChannel(server_ip + ":50051", grpc::InsecureChannelCredentials()));
    client.uploadFile(filename);

    return 0;
}

客户端代码展示了如何创建一个gRPC客户端，如何打开文件，如何将文件切割成块，并且如何将这些块通过网络发送到服务端。

3.2 gRPC服务端实现

服务端的实现则负责接收来自客户端的数据块，并将其写入到服务器上的文件中。服务端代码如下：

#include <iostream>
#include <string>
#include <fstream>
#include <grpcpp/grpcpp.h>
#include <grpcpp/server.h>
#include <grpcpp/server_builder.h>
#include <grpcpp/server_context.h>
#include <grpcpp/security/server_credentials.h>
#include "transferfile.grpc.pb.h"
#include <sys/resource.h>
#include <unistd.h>
#include <chrono>

using grpc::Server;
using grpc::ServerBuilder;
using grpc::ServerContext;
using grpc::ServerReader;
using grpc::Status;
using transferfile::FileChunk;
using transferfile::FileUploadStatus;
using transferfile::TransferFile;
using namespace std::chrono;

#define CHUNK_SIZE 3 * 1024 * 1024 // 3MB

class TransferServiceImpl final : public TransferFile::Service
{
public:
    Status UploadFile(ServerContext *context, ServerReader<FileChunk> *reader, FileUploadStatus *status) override;

protected:
    void printUsage(const struct rusage &start, const struct rusage &end)
    {
        std::cout << "CPU Usage: User time: " << (end.ru_utime.tv_sec - start.ru_utime.tv_sec) + (end.ru_utime.tv_usec - start.ru_utime.tv_usec) / 1000000.0
                  << "s, System time: " << (end.ru_stime.tv_sec - start.ru_stime.tv_sec) + (end.ru_stime.tv_usec - start.ru_stime.tv_usec) / 1000000.0 << "s" << std::endl;
        std::cout << "Max resident set size: " << (end.ru_maxrss - start.ru_maxrss) << " KB" << std::endl;
    }
};

Status TransferServiceImpl::UploadFile(ServerContext *context, ServerReader<FileChunk> *reader, FileUploadStatus *status)
{
    FileChunk chunk;
    std::ofstream outfile;
    const char *data;

    struct rusage usage_start, usage_end;
    getrusage(RUSAGE_SELF, &usage_start);

    auto start = high_resolution_clock::now();

    outfile.open("output.bin", std::ios::binary | std::ios::out | std::ios::trunc);
    if (!outfile.is_open())
    {
        std::cerr << "Error: Failed to open file" << std::endl;
        return Status::CANCELLED;
    }

    while (reader->Read(&chunk))
    {
        data = chunk.buffer().c_str();
        outfile.write(data, chunk.buffer().length());
    }

    long pos = outfile.tellp();
    status->set_length(pos);
    outfile.close();

    auto end = high_resolution_clock::now();
    getrusage(RUSAGE_SELF, &usage_end);
    auto duration = duration_cast<milliseconds>(end - start);

    printUsage(usage_start, usage_end);
    std::cout << "Total transmission time: " << duration.count() << " ms" << std::endl;
    std::cout << "Total data transmitted: " << pos << " bytes" << std::endl;
    std::cout << "Transmission rate: " << (pos * 1000.0 / duration.count()) / 1024 / 1024 << " MB/s" << std::endl;

    return Status::OK;
}

void RunServer()
{
    std::string server_address("0.0.0.0:50051");
    TransferServiceImpl service;

    ServerBuilder builder;
    builder.AddListeningPort(server_address, grpc::InsecureServerCredentials());
    builder.RegisterService(&service);

    std::unique_ptr<Server> server(builder.BuildAndStart());
    std::cout << "Server listening on " << server_address << std::endl;
    server->Wait();
}

int main(int argc, char **argv)
{
    RunServer();
    return 0;
}

服务端代码展示了如何创建一个gRPC服务端，如何接收客户端发送的数据块，以及如何将这些数据块写入到磁盘文件中。

3.3 TCP客户端实现

功能同gRPC客户端

#include <iostream>
#include <fstream>
#include <sys/socket.h>
#include <arpa/inet.h>
#include <unistd.h>
#include <string>
#include <cstring>

int main(int argc, char *argv[])
{
    const char *server_ip = "127.0.0.1";
    int server_port = 8888;
    std::string file_path = argv[1];

    // Create socket
    int sock = socket(AF_INET, SOCK_STREAM, 0);
    if (sock < 0)
    {
        std::cerr << "Socket creation failed." << std::endl;
        return 1;
    }

    // Define the server address
    struct sockaddr_in server_address;
    memset(&server_address, 0, sizeof(server_address));
    server_address.sin_family = AF_INET;
    server_address.sin_port = htons(server_port);
    inet_pton(AF_INET, server_ip, &server_address.sin_addr);

    // Connect to the server
    if (connect(sock, (struct sockaddr *)&server_address, sizeof(server_address)) < 0)
    {
        std::cerr << "Connection to server failed." << std::endl;
        close(sock);
        return 1;
    }

    std::ifstream file(file_path, std::ios::binary);
    if (!file.is_open())
    {
        std::cerr << "Failed to open file: " << file_path << std::endl;
        close(sock);
        return 1;
    }

    // Send file contents
    char *buffer = new char[1024];
    while (file.read(buffer, sizeof(buffer)) || file.gcount())
    {
        if (send(sock, buffer, file.gcount(), 0) < 0)
        {
            std::cerr << "Failed to send data." << std::endl;
            break;
        }
    }

    delete[] buffer;

    std::cout << "File sent successfully." << std::endl;

    file.close();
    close(sock);

    return 0;
}

3.4 TCP服务端实现

功能同gRPC服务端

#include <iostream>
#include <fstream>
#include <sys/socket.h>
#include <arpa/inet.h>
#include <unistd.h>
#include <cstring>
#include <string>
#include <sys/resource.h>
#include <unistd.h>
#include <chrono>

#define CHUNK_SIZE 3 * 1024 * 1024 // 3MB

using namespace std::chrono;

void printUsage(const struct rusage &start, const struct rusage &end)
{
    std::cout << "CPU Usage: User time: " << (end.ru_utime.tv_sec - start.ru_utime.tv_sec) + (end.ru_utime.tv_usec - start.ru_utime.tv_usec) / 1000000.0
              << "s, System time: " << (end.ru_stime.tv_sec - start.ru_stime.tv_sec) + (end.ru_stime.tv_usec - start.ru_stime.tv_usec) / 1000000.0 << "s" << std::endl;
    std::cout << "Max resident set size: " << (end.ru_maxrss - start.ru_maxrss) << " KB" << std::endl;
}

int main()
{
    int server_port = 8888;
    const char *output_file = "output_received_file";

    // Create socket
    int server_sock = socket(AF_INET, SOCK_STREAM, 0);
    if (server_sock < 0)
    {
        std::cerr << "Socket creation failed." << std::endl;
        return 1;
    }

    // Bind socket to IP / port
    struct sockaddr_in server_address;
    memset(&server_address, 0, sizeof(server_address));
    server_address.sin_family = AF_INET;
    server_address.sin_port = htons(server_port);
    server_address.sin_addr.s_addr = INADDR_ANY;

    if (bind(server_sock, (struct sockaddr *)&server_address, sizeof(server_address)) < 0)
    {
        std::cerr << "Bind failed." << std::endl;
        close(server_sock);
        return 1;
    }

    // Listen
    if (listen(server_sock, 10) < 0)
    {
        std::cerr << "Listen failed." << std::endl;
        close(server_sock);
        return 1;
    }

    std::cout << "Server is listening on port " << server_port << std::endl;

    // Accept connection
    struct sockaddr_in client_address;
    socklen_t client_len = sizeof(client_address);
    int client_sock = accept(server_sock, (struct sockaddr *)&client_address, &client_len);
    if (client_sock < 0)
    {
        std::cerr << "Accept failed." << std::endl;
        close(server_sock);
        return 1;
    }

    struct rusage usage_start, usage_end;
    getrusage(RUSAGE_SELF, &usage_start);

    auto start = high_resolution_clock::now();

    std::ofstream file(output_file, std::ios::binary);
    if (!file.is_open())
    {
        std::cerr << "Failed to open file for writing." << std::endl;
        close(client_sock);
        close(server_sock);
        return 1;
    }

    // Receive data
    char *buffer = new char[CHUNK_SIZE];
    int bytes_received;
    while ((bytes_received = recv(client_sock, buffer, sizeof(buffer), 0)) > 0)
    {
        file.write(buffer, bytes_received);
    }

    if (bytes_received < 0)
    {
        std::cerr << "Error in recv()." << std::endl;
    }
    else
    {
        std::cout << "File received successfully." << std::endl;
        long pos = file.tellp();
        auto end = high_resolution_clock::now();
        getrusage(RUSAGE_SELF, &usage_end);
        auto duration = duration_cast<milliseconds>(end - start);

        printUsage(usage_start, usage_end);
        std::cout << "Total transmission time: " << duration.count() << " ms" << std::endl;
        std::cout << "Total data transmitted: " << pos << " bytes" << std::endl;
        std::cout << "Transmission rate: " << (pos * 1000.0 / duration.count()) / 1024 / 1024 << " MB/s" << std::endl;
    }

    delete[] buffer;

    file.close();
    close(client_sock);
    close(server_sock);
    return 0;
}

4. 性能测试与分析

为了验证gRPC与Protobuf的效率，我设置了一个基准测试，比较使用gRPC和传统TCP直接传输大文件的性能差异。

4.1 测试方法

测试方法包括：

准备一定大小的测试文件，这里随机生成了2GB的文件。（fallocate -l 2G 2GBfile.txt）
分别使用gRPC和TCP传输此文件，记录所需的总时间和CPU、内存等资源的使用情况。
重复测试，确保数据的准确性。

4.1.1 gPRC测试结果

4.1.2 TCP测试结果

4.1.3 多次测试结果

	User time	System time	Max resident set size	Transmission time	Transmission rate
gRPC	28.5267 s	17.2359 s	18240 KB	18.265 s	112.127 MB/s
TCP	43.2163 s	143.779 s	0 KB	187.408 s	10.928 MB/s
gRPC	29.3631 s	17.477 s	17724 KB	18.312 s	111.839 MB/s
TCP	42.2783 s	140.658 s	0 KB	183.123 s	11.1837 MB/s
gRPC	28.8431 s	17.7514 s	16852 KB	18.691 s	109.571 MB/s
TCP	43.4233 s	144.232 s	0 KB	189.021 s	10.834 MB/s
gRPC	37.5632 s	19.1164 s	15424 KB	18.589 s	110.173 MB/s
TCP	40.2059 s	140.3 s	0 KB	180.772 s	11.3292 MB/s
gRPC	36.712 s	18.1812 s	14016 KB	18.274 s	112.072 MB/s
TCP	37.9126 s	144.276 s	0 KB	182.392 s	11.2286 MB/s

4.2 性能比较结果

4.2.1 CPU 使用

gRPC在CPU使用上明显低于传统的TCP socket方法。这主要因为gRPC内部使用了更现代的HTTP/2协议，它支持多路复用、服务器推送等高效的数据传输机制，而不需要像TCP那样对每个文件传输任务建立单独的连接。此外，gRPC的实现中可能包含了更优化的数据处理路径，减少了上下文切换和系统调用的开销。

4.2.2 内存复用

Max resident set size（最大常驻集大小）表示进程在内存中占用的最大空间。TCP方式的最大常驻集大小一直是0KB，而gRPC分配的比较多，这是由于gRPC框架本身的内存需求，以及可能的内存缓冲机制，这有助于提高数据处理的速率和效率。

4.2.3 传输时间和速率

gRPC在传输速度上极大超过了TCP socket。这种巨大的差异主要来自于gRPC使用HTTP/2的优势，如头部压缩、二进制帧传输和连接复用。HTTP/2的二进制帧结构使得传输更加高效，并减少了因为文本解析带来的开销。此外，连接复用允许在单一连接上并行交换消息，从而显著提升了数据传输效率，减少了因建立和关闭多个TCP连接所产生的延迟和资源消耗。

测试结果显示，使用gRPC和Protobuf传输大文件在多个方面均优于传统TCP方法：

传输速度: gRPC利用HTTP/2的多路复用功能，可以在一个连接中并行传输多个文件，显著提升了传输效率。
资源利用率: gRPC的传输过程中CPU利用率较低，内存复用率较高，更适合长时间运行的应用场景。
错误处理: gRPC内置的错误处理机制能有效地管理网络问题和数据传输错误，保证数据的完整性。
高效的数据序列化: Protobuf非常高效，生成的数据包体积小，通常比相等的XML小3到10倍。这意味着在网络上传输相同的数据量时，Protobuf需要的带宽更少。
快速序列化与反序列化: Protobuf提供了非常快的数据序列化和反序列化能力，这对于性能要求高的应用尤其重要，可以显著减少数据处理时间。
明确的结构定义: 使用Protobuf可以让数据结构更加清晰和严格，有助于团队内部的沟通和后期的维护。
避免手动解析:与自定义的二进制格式相比，Protobuf避免了手动解析数据的错误和复杂性，因为解析工作是自动化的，由工具链支持。