赵斌的博客

Python 多线程XML-RPC的实现

赵斌 2009年08月15日星期六 01:13 | 3828次浏览 | 0条评论

原文通过blogbus2zeuux工具同步发布:http://antmanler.blogbus.com/logs/43841379.html

文章发布时间:2009/4/10/3/21

Python 多线程XML-RPC的实现

一、引言

用Python的SimpleXMLRPCServer实现的XML-RPC服务器只能是单线程，一个一个的处理请求，对于操作中耗时较长的操作会阻塞其他用户请求，使服务质量下降，因此有没有办法让SimpleXMLRPCServer成为多线程服务器呢？

主要有两种实现方法：修改SimpleXMLRPCServer的实现代码，用ThreadingMixIn或 ForkingMixIn。

二、 修改SimpleXMLRPCServer的实现代码

SimpleXMLRPCServer的实现如下：

class SimpleXMLRPCServer(SocketServer.TCPServer,

SimpleXMLRPCDispatcher):

如果我们将其继承的SocketServer.TCPServer,改为支持线程处理的ThreadingTCPServer，即可将SimpleXMLRPCServer变成多线程的服务器，如下：

class SimpleXMLRPCServer(SocketServer.ThreadingTCPServer,

SimpleXMLRPCDispatcher):

三、 用ThreadingMixIn实现

SocketServer模块中就支持对线程（Threading）或进程（Fork）的支持，典型代码如下：

      
       import
      
       SimpleXMLRPCServer
      
       ,
      
       SocketServer
      
       import
      
       win32api
      
       ,
      
       thread
      
       #The server object
      
       class
      
       Server
      
       :
      
       count
      
       =
      
       0
      
       def
      
       __init__
      
      (
      
       self
      
       ):
      
       pass
      
       def
      
       echoString
      
      (
      
       self
      
       ,
      
       strs
      
       ):
      
       mutex
      
       .
      
       acquire
      
      ()
      
       #用mutex锁住数据
      
       Server
      
       .
      
       count
      
       +=
      
       1
      
       #更改静态数据
      
       t
      
       =
      
       time
      
       .
      
       strftime
      
      (
      
       "Serve at %I:%M:%S Id is "
      
      )
      
       +
      
       str
      
      (
      
       Server
      
       .
      
       count
      
      )
      
       print
      
       "Serve Id "
      
       +
      
       str
      
      (
      
       Server
      
       .
      
       count
      
      )
      
       mutex
      
       .
      
       release
      
      ()
      
       #释放锁
      
       #win32api.Sleep(20000)
      
       return
      
       t
      
       #多线程实现
      
       class
      
       RPCThreading
      
      (
      
       SocketServer
      
       .
      
       ThreadingMixIn
      
       ,
      
       SimpleXMLRPCServer
      
       .
      
       SimpleXMLRPCServer
      
       ):
      
       pass
      
       global
      
       mutex
      
       mutex
      
       =
      
       thread
      
       .
      
       allocate_lock
      
      ()
      
       server_object
      
       =
      
       Server
      
      ()
      
       server
      
       =
      
       RPCThreading
      
      ((
      
       "localhost"
      
       ,
      
       8888
      
      ))
      
       server
      
       .
      
       register_instance
      
      (
      
       server_object
      
      )
      
       #Go into the main listener loop
      
       print
      
       "Listening on 8888"
      
       server
      
       .
      
       serve_forever
      
      ()

通过上面的代码也可以实现XML-RPC的多线程支持而不用修改Python的 SimpleXMLRPCServer库。

转:下面顺便转个压缩发送的:

compression. (Disclaimer : I'm not a python expert in any way, just sharing...)

xmlrpc is great, easy, dynamic but is not effecient as binary methods (see "The Python Web services developer: Messaging technologies compared" at http://www.ibm.com/developerworks/library/ws-pyth9/).

For example, it can take minutes to transfer 100MB of data between client and server.

Therefore, I added compression, using http://aspn.activestate.com/ASPN/Cookbook/Python/Recipe/355486.

This shortened the transfer time to 15 seconds. Example server function (returns an arbitrary chunk of bytes, possibly with random values):

    
     from
    
     array
    
     import
    
     array
    
     from
    
     random
    
     import
    
     randint
    
     import
    
     zlib
    
     ,
    
     base64
    
     def
    
     get_data
    
    (
    
     self
    
     ,
    
     default_byte_value
    
     =
    
     0x00
    
     ,
    
     num_bytes
    
     =
    
     1024
    
     ,
    
     use_rand
    
     =
    
     False
    
     ):
    
     data
    
     =
    
     array
    
    (
    
     'B'
    
     ,
    
     [
    
     default_byte_value
    
     ]
    
     *
    
     num_bytes
    
    )
    
     if
    
    (
    
     use_rand
    
     ):
    
     for
    
     i
    
     in
    
     range
    
    (
    
     num_bytes
    
     ):
    
     data
    
     [
    
     i
    
     ]
    
     =
    
     randint
    
    (
    
     0
    
     ,
    
     255
    
    )
    
     data
    
     ,
    
     csum
    
     =
    
     asciiCompress
    
    (
    
     buffer
    
    (
    
     data
    
    ))
    
     return
    
     data

Example client side (request a 1MB chunk of bytes using 0x11 as default value for each byte (non random)).

    
     data
    
     =
    
     server
    
     .
    
     get_data
    
    (
    
     0x11
    
     ,
    
     1
    
     *
    
     10
    
     **
    
     6
    
     ,
    
     False
    
    )
    
     data
    
     ,
    
     csum
    
     =
    
     asciiDecompress
    
    (
    
     data
    
    )
    
     arr
    
     =
    
     array
    
    (
    
     'B'
    
     ,
    
     data
    
    )

Note that we originate from a python array at the server and end up with an array at the client.

Just my 2 cents :)

分享添加到桌面

暂时没有评论