Commit fddfd1eb authored by 蔡宇东's avatar 蔡宇东
Browse files

#346 update config APIs to support build index with multiple GPUs

parent a47b7284

core/conf/server_cpu_config.template

+3 −3
Original line number Diff line number Diff line
@@ -27,7 +27,6 @@ metric_config: 
    port: 8080                      # port prometheus uses to fetch metrics, must in range [1025, 65534]



cache_config:



  cpu_cache_capacity: 16            # GB, CPU memory used for cache, must be a positive integer

  cpu_cache_threshold: 0.85         # percentage of data that will be kept when cache cleanup is triggered, must be in range (0.0, 1.0]

  cache_insert_data: false          # whether to load inserted data into cache, must be a boolean
@@ -38,6 +37,7 @@ engine_config: 
  gpu_search_threshold: 1000        # threshold beyond which the search computation is executed on GPUs only



resource_config:

  search_resources:                 # define the device used for search computation

  search_resources:                 # define the devices used for search computation, must be in format: cpu or gpux

    - cpu

  index_build_resources:            # define the devices used for index building, must be in format: cpu or gpux

    - cpu
 
  index_build_device: cpu           # CPU used for building index

core/conf/server_gpu_config.template

+2 −1
Original line number Diff line number Diff line
@@ -42,4 +42,5 @@ resource_config: 
  search_resources:                 # define the devices used for search computation, must be in format: cpu or gpux

    - cpu

    - gpu0

  index_build_device: gpu0          # CPU / GPU used for building index, must be in format: cpu or gpux
 
  index_build_resources:            # define the devices used for index building, must be in format: cpu or gpux

    - gpu0
 
 No newline at end of file

core/src/server/Config.cpp

+46 −38
Original line number Diff line number Diff line
@@ -215,8 +215,8 @@ Config::ValidateConfig() { 
        return s;

    }



    int32_t resource_index_build_device;

    s = GetResourceConfigIndexBuildDevice(resource_index_build_device);

    std::vector<std::string> index_build_resources;

    s = GetResourceConfigIndexBuildResources(index_build_resources);

    if (!s.ok()) {

        return s;

    }
@@ -351,7 +351,7 @@ Config::ResetDefaultConfig() { 
        return s;

    }



    s = SetResourceConfigIndexBuildDevice(CONFIG_RESOURCE_INDEX_BUILD_DEVICE_DEFAULT);

    s = SetResourceConfigIndexBuildResources(CONFIG_RESOURCE_INDEX_BUILD_RESOURCES_DEFAULT);

    if (!s.ok()) {

        return s;

    }
@@ -599,13 +599,18 @@ Config::CheckCacheConfigGpuCacheCapacity(const std::string& value) { 
        return Status(SERVER_INVALID_ARGUMENT, msg);

    } else {

        uint64_t gpu_cache_capacity = std::stoi(value) * GB;

        int device_id;

        Status s = GetResourceConfigIndexBuildDevice(device_id);

        std::vector<std::string> resources;

        Status s = GetResourceConfigIndexBuildResources(resources);

        if (!s.ok()) {

            return s;

        }



        size_t gpu_memory;

        for (auto& resource : resources) {

            if (resource == "cpu") {

                continue;

            }

            int32_t device_id = std::stoi(resource.substr(3));

            if (!ValidationUtil::GetGpuMemory(device_id, gpu_memory).ok()) {

                std::string msg = "Fail to get GPU memory for GPU device: " + std::to_string(device_id);

                return Status(SERVER_UNEXPECTED_ERROR, msg);
@@ -617,6 +622,7 @@ Config::CheckCacheConfigGpuCacheCapacity(const std::string& value) { 
                std::cerr << "Warning: gpu cache capacity value is too big" << std::endl;

            }

        }

    }

    return Status::OK();

}
@@ -745,11 +751,19 @@ Config::CheckResourceConfigSearchResources(const std::vector<std::string>& value 
}



Status

Config::CheckResourceConfigIndexBuildDevice(const std::string& value) {

    auto status = CheckResource(value);

Config::CheckResourceConfigIndexBuildResources(const std::vector<std::string>& value) {

    if (value.empty()) {

        std::string msg =

            "Invalid build index resource. "

            "Possible reason: resource_config.build_index_resources is empty.";

        return Status(SERVER_INVALID_ARGUMENT, msg);

    }

    for (auto& resource : value) {

        auto status = CheckResource(resource);

        if (!status.ok()) {

            return Status(SERVER_INVALID_ARGUMENT, status.message());

        }

    }

    return Status::OK();

}
@@ -1030,27 +1044,18 @@ Status 
Config::GetResourceConfigSearchResources(std::vector<std::string>& value) {

    std::string str =

        GetConfigSequenceStr(CONFIG_RESOURCE, CONFIG_RESOURCE_SEARCH_RESOURCES,

                             CONFIG_RESOURCE_SEARCH_RESOURCES_DELIMITER, CONFIG_RESOURCE_SEARCH_RESOURCES_DEFAULT);

    server::StringHelpFunctions::SplitStringByDelimeter(str, CONFIG_RESOURCE_SEARCH_RESOURCES_DELIMITER, value);

                             CONFIG_RESOURCE_RESOURCES_DELIMITER, CONFIG_RESOURCE_SEARCH_RESOURCES_DEFAULT);

    server::StringHelpFunctions::SplitStringByDelimeter(str, CONFIG_RESOURCE_RESOURCES_DELIMITER, value);

    return CheckResourceConfigSearchResources(value);

}



Status

Config::GetResourceConfigIndexBuildDevice(int32_t& value) {

Config::GetResourceConfigIndexBuildResources(std::vector<std::string>& value) {

    std::string str =

        GetConfigStr(CONFIG_RESOURCE, CONFIG_RESOURCE_INDEX_BUILD_DEVICE, CONFIG_RESOURCE_INDEX_BUILD_DEVICE_DEFAULT);

    Status s = CheckResourceConfigIndexBuildDevice(str);

    if (!s.ok()) {

        return s;

    }



    if (str == "cpu") {

        value = CPU_DEVICE_ID;

    } else {

        value = std::stoi(str.substr(3));

    }



    return Status::OK();

        GetConfigSequenceStr(CONFIG_RESOURCE, CONFIG_RESOURCE_INDEX_BUILD_RESOURCES,

                        CONFIG_RESOURCE_RESOURCES_DELIMITER, CONFIG_RESOURCE_INDEX_BUILD_RESOURCES_DEFAULT);

    server::StringHelpFunctions::SplitStringByDelimeter(str, CONFIG_RESOURCE_RESOURCES_DELIMITER, value);

    return CheckResourceConfigIndexBuildResources(value);

}



///////////////////////////////////////////////////////////////////////////////
@@ -1305,7 +1310,7 @@ Config::SetResourceConfigMode(const std::string& value) { 
Status

Config::SetResourceConfigSearchResources(const std::string& value) {

    std::vector<std::string> res_vec;

    server::StringHelpFunctions::SplitStringByDelimeter(value, CONFIG_RESOURCE_SEARCH_RESOURCES_DELIMITER, res_vec);

    server::StringHelpFunctions::SplitStringByDelimeter(value, CONFIG_RESOURCE_RESOURCES_DELIMITER, res_vec);



    Status s = CheckResourceConfigSearchResources(res_vec);

    if (!s.ok()) {
@@ -1317,13 +1322,16 @@ Config::SetResourceConfigSearchResources(const std::string& value) { 
}



Status

Config::SetResourceConfigIndexBuildDevice(const std::string& value) {

    Status s = CheckResourceConfigIndexBuildDevice(value);

Config::SetResourceConfigIndexBuildResources(const std::string &value) {

    std::vector<std::string> res_vec;

    server::StringHelpFunctions::SplitStringByDelimeter(value, CONFIG_RESOURCE_RESOURCES_DELIMITER, res_vec);



    Status s = CheckResourceConfigIndexBuildResources(res_vec);

    if (!s.ok()) {

        return s;

    }



    SetConfigValueInMem(CONFIG_RESOURCE, CONFIG_RESOURCE_INDEX_BUILD_DEVICE, value);

    SetConfigValueInMem(CONFIG_RESOURCE, CONFIG_RESOURCE_INDEX_BUILD_RESOURCES, value);

    return Status::OK();

}

core/src/server/Config.h

+7 −9
Original line number Diff line number Diff line
@@ -91,20 +91,18 @@ static const char* CONFIG_ENGINE_GPU_SEARCH_THRESHOLD_DEFAULT = "1000"; 
static const char* CONFIG_RESOURCE = "resource_config";

static const char* CONFIG_RESOURCE_MODE = "mode";

static const char* CONFIG_RESOURCE_MODE_DEFAULT = "simple";

static const char* CONFIG_RESOURCE_RESOURCES_DELIMITER = ",";

static const char* CONFIG_RESOURCE_SEARCH_RESOURCES = "search_resources";

static const char* CONFIG_RESOURCE_SEARCH_RESOURCES_DELIMITER = ",";



#ifdef MILVUS_CPU_VERSION

static const char* CONFIG_RESOURCE_SEARCH_RESOURCES_DEFAULT = "cpu";

#else

static const char* CONFIG_RESOURCE_SEARCH_RESOURCES_DEFAULT = "cpu,gpu0";

#endif



static const char* CONFIG_RESOURCE_INDEX_BUILD_DEVICE = "index_build_device";

static const char* CONFIG_RESOURCE_INDEX_BUILD_RESOURCES = "index_build_resources";

#ifdef MILVUS_CPU_VERSION

static const char* CONFIG_RESOURCE_INDEX_BUILD_DEVICE_DEFAULT = "cpu";

static const char* CONFIG_RESOURCE_INDEX_BUILD_RESOURCES_DEFAULT = "cpu";

#else

static const char* CONFIG_RESOURCE_INDEX_BUILD_DEVICE_DEFAULT = "gpu0";

static const char* CONFIG_RESOURCE_INDEX_BUILD_RESOURCES_DEFAULT = "gpu0";

#endif

const int32_t CPU_DEVICE_ID = -1;
@@ -190,7 +188,7 @@ class Config { 
    Status

    CheckResourceConfigSearchResources(const std::vector<std::string>& value);

    Status

    CheckResourceConfigIndexBuildDevice(const std::string& value);

    CheckResourceConfigIndexBuildResources(const std::vector<std::string>& value);



    std::string

    GetConfigStr(const std::string& parent_key, const std::string& child_key, const std::string& default_value = "");
@@ -259,7 +257,7 @@ class Config { 
    Status

    GetResourceConfigSearchResources(std::vector<std::string>& value);

    Status

    GetResourceConfigIndexBuildDevice(int32_t& value);

    GetResourceConfigIndexBuildResources(std::vector<std::string>& value);



 public:

    /* server config */
@@ -320,7 +318,7 @@ class Config { 
    Status

    SetResourceConfigSearchResources(const std::string& value);

    Status

    SetResourceConfigIndexBuildDevice(const std::string& value);

    SetResourceConfigIndexBuildResources(const std::string& value);



 private:

    std::unordered_map<std::string, std::unordered_map<std::string, std::string>> config_map_;

core/src/utils/ValidationUtil.cpp

+2 −2
Original line number Diff line number Diff line
@@ -182,7 +182,7 @@ ValidationUtil::ValidatePartitionTags(const std::vector<std::string>& partition_ 
}



Status

ValidationUtil::ValidateGpuIndex(uint32_t gpu_index) {

ValidationUtil::ValidateGpuIndex(int32_t gpu_index) {

#ifdef MILVUS_GPU_VERSION

    int num_devices = 0;

    auto cuda_err = cudaGetDeviceCount(&num_devices);
@@ -203,7 +203,7 @@ ValidationUtil::ValidateGpuIndex(uint32_t gpu_index) { 
}



Status

ValidationUtil::GetGpuMemory(uint32_t gpu_index, size_t& memory) {

ValidationUtil::GetGpuMemory(int32_t gpu_index, size_t& memory) {

#ifdef MILVUS_GPU_VERSION



    cudaDeviceProp deviceProp;