FlatBuffers與protobuf效能比較

標籤：des   class   http   tar   com   get   

    FlatBuffers發布時，順便也公布了它的效能資料，具體資料請見Benchmark。

    它的測試案例由以下資料構成"a set of about 10 objects containing an array, 4 strings, and a large variety of int/float scalar values of all sizes, meant to be representative of game data, e.g. a scene format."

   我感覺這樣測試如同兒戲，便自己設計了一個測試案例，主要關注CPU計算時間和記憶體空間佔用兩個指標，參考對象是protobuf。

   測試案例為：序列化一個通訊錄personal_info_list（table），通訊錄可以認為是有每個人的資訊(personal_info)的集合。每個人資訊personal_info(table)有：個人id(uint)、名字(string)、年齡(byte)、性別(enum, byte)和電話號碼(ulong)。本來我想用struct表示personal_info(table)，但是struct不允許有數組或string成員，無奈我用table描述它了。相應的idl檔案如下：

 

////////////////////////////////////////////////////////// FILE     : tellist.fbs//// DESC     : basic message for msg-center//// AUTHOR   : v 0.1 written by Alex Stocks on June 22, 2014//// LICENCE  ://// MOD      :////////////////////////////////////////////////////////namespace as.tellist;enum GENDER_TYPE : byte{MALE= 0,FEMALE= 1,OTHER= 2}table personal_info{id : uint;name : string;age : byte;gender : GENDER_TYPE;phone_num : ulong;}table personal_info_list{info : [personal_info];}root_type personal_info_list;

   因為要以protobuf做效能參考，列出protobuf的idl檔案如下：    

 

////////////////////////////////////////////////////////// FILE     : tellist.proto//// DESC     : basic message for msg-center//// AUTHOR   : v 0.1 written by Alex Stocks on June 22, 2014//// LICENCE  ://// MOD      :////////////////////////////////////////////////////////package as.tellist;enum gender_type{MALE= 0;FEMALE= 1;OTHER= 2;}message personal_info{optional uint32id = 1;optional stringname = 2;optional uint32age = 3;optional gender_typegender = 4;optional uint64phone_num = 5;}message personal_info_list{repeated personal_infoinfo = 1;}

 

    在記憶體中構造37個personal_info對象，並序列化之，重複這個過程100萬次。

    測試結果如下：

 

測試環境：12Core Intel(R) Xeon(R) CPU E5-2620 0 @ 2.00GHz free             total       used       free     shared    buffers     cachedMem:      66081944   65831100     250844          0     182240   46903452-/+ buffers/cache:   18745408   47336536Swap:       975864     724648     251216protobuf三次測試結果：./bin/tellist_test loop = 1000000, time diff = 14283msbuf size:841bin/tellist_test loop = 1000000, time diff = 14096msbuf size:841bin/tellist_test loop = 1000000, time diff = 14229msbuf size:841佔用記憶體空間841Byte，平均運算時間42608ms / 3 = 14202.7msflatbuffers三次測試結果：bin/tellist_test loop = 1000000, time diff = 11694msbuf size:1712 bin/tellist_test loop = 1000000, time diff = 11710msbuf size:1712bin/tellist_test loop = 1000000, time diff = 11774msbuf size:1712佔用記憶體空間1712Byte，平均運算時間35178ms / 3 = 11726ms測試環境：1 Core Intel(R) Core(TM) i5-3210M CPU @ 2.50GHzMEM             total       used       free     shared    buffers     cachedMem:        753932     432672     321260          0      89236     258052-/+ buffers/cache:      85384     668548Swap:      1324028          0    1324028protobuf三次測試結果：bin/tellist_test loop = 1000000, time diff = 12779msbuf size:841bin/tellist_test loop = 1000000, time diff = 13475msbuf size:841bin/tellist_test loop = 1000000, time diff = 12604msbuf size:841佔用記憶體空間841Byte，平均運算時間38858ms / 3 = 12952.7msflatbuffers三次測試結果：bin/tellist_test loop = 1000000, time diff = 9424msbuf size:1712 bin/tellist_test loop = 1000000, time diff = 9277msbuf size:1712bin/tellist_test loop = 1000000, time diff = 9265msbuf size:1712info vecotor size:37, its right size:37佔用記憶體空間1712Byte，平均運算時間28036ms / 3 = 9345ms

    從以上資料看出，在記憶體空間佔用這個指標上，FlatBuffers佔用的記憶體空間比protobuf多了兩倍，而二者的cpu計算時間雖然相差3000ms左右，但考慮到測試用了100萬次，二者每次計算時間幾乎沒有差別。

    從以上測試資料來看，FlatBuffers的效能並不如它吹噓的那麼好，個人稍有點失望。FB大量使用了C++11的文法，其從idl產生的程式碼介面也不如protubuf友好。不過相比使用protobuf時的一堆標頭檔和佔18M之多的lib庫，FlatBuffers僅僅一個"flatbuffers/flatbuffers.h"就足夠了。

    測試程式已經上傳到百度網盤，點擊這個連結即可下載。歡迎各位的批評意見。