saf application framework
1:基于google protobuf的rpc框架;
2:内置简单的http server用于查看系统状态;
3:模块化的管理,只用在lib中增加动态库就可以达到增加rpc service的目的;
4:sdk支持同步和异步调用模式、断线自动重连、调用超时设置;(查看example/rpcping/client.cc)
+--------------+
+-----------+ +------------+ /| |
| | | | / | Handle Thread| --->-
|Net Thread \----->| Decode |\ / +--------------+ |
+-----------+\ +------------+ \ +-----------------+ / +--------------+ |
\ \>| |/ | | |
\ />| queue +-----+ Handle Thread+---->+
+-----------+ \ +------------+ / +-----------------+\ +--------------+ |
| | \ | |/ \ +--------------+ |
|Net Thread +----+>+ Decode / \ | | |
+-----------+ \+------------+ \ | Handle Thread| |
\ \ +---------+----+ |
\ \ | |
\ \ | |
\ +--\---------+ | |
\ | |<-------------------------------------+ |
\ | Encode | |
\ +------------+ |
\ +------------+ |
\| | |
\ Encode +<------------------------------------------------+
+------------+
它依赖protobuf公共库,所以先安装protobuf3.0及以上版本，安装protobuf参照：
https://github.com/google/protobuf
1:下载源码
git clone --recursive https://github.com/sails/saf.git
2:编译:
make depends
make;make deploy;
3:编译测试样例:
make examples
4:配置测试程序(conf/saf.json):
"modules" : [{"name":"test", "path":"../lib/service_test.so"}]
5:运行测试,比如运行echo_sync:
./bin/saf
./example/echo_sync/client
编译选项：-O0, 四核八线程cpu 服务端和客户端都运行在这台电脑上,每个客户端同步调用10w次
| net thread | handle thread | clients | times | tps |
|------------+---------------+---------+--------+--------|
| 1 | 1 | 1 | 9.097 | 11000 |
| 1 | 1 | 30 | 37.986 | 79000 |
| 2 | 1 | 30 | 33.872 | 88000 |
| 2 | 2 | 30 | 22.302 | 135000 |
| 2 | 4 | 60 | 37.841 | 158000 |
| 3 | 3 | 60 | 34.312 | 175000 |
| 4 | 4 | 100 | 50.117 | 200000 |
| 4 | 8 | 100 | 48.383 | 207000 |
通过测试可知，在一台普通的四核八线程服务器上，基本上能达到20w/s的速度 (单独测试protobuf时，单线程处理10w次序列化和反序列化用时2s)
内置http服务器的默认绑定在端口8001上,所以可以通过输入localhost:8001/stat来查看统计信息:
