Everyone knows theasyncio module in python schedules all coroutines in a single thread. That means it helps you code easier, and you can't gain any performance from it.
But what is the performance of python'sasyncio module? How fast can it run compared to traditionalgevent and nativeepoll ?
Requirements
pip3installhiredis geventAn optional packageuvloop can also be install if working on Linux:
pip3installuvloopSource
The following code uses bothasyncio andgevent to simulate a redis server on port 5000, 5001, and 5002.
They can be tested withredis-benchmark.
content ofecho_bench_gevent.py:
importsysimportgeventimportgevent.monkeyimporthiredisfromgevent.serverimportStreamServergevent.monkey.patch_all()d={}defprocess(req):# only support get/setcmd=req[0].lower()ifcmd==b'set':d[req[1]]=req[2]returnb"+OK\r\n"elifcmd==b'get':v=d.get(req[1])ifvisNone:returnb'$-1\r\n'else:returnb'$1\r\n1\r\n'else:print(cmd)raiseNotImplementedError()returnb''defhandle(sock,addr):reader=hiredis.Reader()whileTrue:buf=sock.recv(4096)ifnotbuf:returnreader.feed(buf)whileTrue:req=reader.gets()ifnotreq:breaksock.sendall(process(req))return0print('serving on 0.0.0.0:5000')server=StreamServer(('0.0.0.0',5000),handle)server.serve_forever()content ofecho_bench_asyncio.py:
importasyncioimporthiredisd={}defprocess(req):cmd=req[0].lower()ifcmd==b'set':d[req[1]]=req[2]returnb"+OK\r\n"elifcmd==b'get':v=d.get(req[1])ifvisNone:returnb'$-1\r\n'else:returnb'$1\r\n1\r\n'elifcmd==b'config':returnb'-ERROR\r\n'else:returnb'-ERROR\r\n'returnb''asyncdefecho_server(reader,writer):hireader=hiredis.Reader()whileTrue:s=awaitreader.read(4096)ifnots:breakhireader.feed(s)whileTrue:req=hireader.gets()ifnotreq:breakres=process(req)writer.write(res)awaitwriter.drain()return0asyncdefmain():server=awaitasyncio.start_server(echo_server,'0.0.0.0',5001)print('serving on {}'.format(server.sockets[0].getsockname()))awaitserver.serve_forever()return0asyncio.run(main())content ofecho_bench_asyncio_uvloop.py:
importasyncioimporthiredisd={}defprocess(req):cmd=req[0].lower()ifcmd==b'set':d[req[1]]=req[2]returnb"+OK\r\n"elifcmd==b'get':v=d.get(req[1])ifvisNone:returnb'$-1\r\n'else:returnb'$1\r\n1\r\n'elifcmd==b'config':returnb'-ERROR\r\n'else:returnb'-ERROR\r\n'returnb''asyncdefecho_server(reader,writer):hireader=hiredis.Reader()whileTrue:s=awaitreader.read(4096)ifnots:breakhireader.feed(s)whileTrue:req=hireader.gets()ifnotreq:breakres=process(req)writer.write(res)awaitwriter.drain()return0asyncdefmain():server=awaitasyncio.start_server(echo_server,'0.0.0.0',5002)print('serving on {}'.format(server.sockets[0].getsockname()))awaitserver.serve_forever()return0try:importuvloopuvloop.install()print('uvloop is enabled')exceptImportError:print('uvloop is not available')asyncio.run(main())Start servers
python3 echo_bench_gevent.py# will listen on port 5000python3 echo_bench_asyncio.py# will listen on port 5001python3 echo_bench_asyncio_uvloop.py# will listen on port 5002Test
redis-benchmark-p 5000-t get-n 100000-r 100000000redis-benchmark-p 5001-t get-n 100000-r 100000000redis-benchmark-p 5002-t get-n 100000-r 100000000Result
| Mode | Python 3.9 | Python 3.11 |
|---|---|---|
| gevent | 34281.80 requests / second | 32258.07 requests / second |
| asyncio | 40144.52 requests / second | 51652.89 requests / second |
| asyncio + uvloop | 64102.57 requests / second | 66577.90 requests / second |
Native epoll
Redis is implemented withepoll in C, and we can test redis directly:
redis-benchmark-p 6379-t get-n 100000-r 100000000Output:
75244.55 requests per secondConclusion
asynciois 50% faster thangeventin Python 3.11asynciocan run twice as fast asgeventwithuvloop.asynciocan go up to 68% of a native epoll program.asynciocan go up to 88% of a native epoll program withuvloop.
Top comments(2)
Subscribe
Thank you for the nice write-up. Using asyncio + uvloop + transports/protocols is even faster and on par with the benchmark against native Redis on my machine:
importasyncioimporthiredisd={}defprocess(req):cmd=req[0].lower()ifcmd==b'set':d[req[1]]=req[2]returnb"+OK\r\n"elifcmd==b'get':v=d.get(req[1])ifvisNone:returnb'$-1\r\n'else:returnb'$1\r\n1\r\n'elifcmd==b'config':returnb'-ERROR\r\n'else:returnb'-ERROR\r\n'returnb''classRedisServerProtocol(asyncio.Protocol):defconnection_made(self,transport):self.transport=transportself.hireader=hiredis.Reader()defdata_received(self,data):self.hireader.feed(data)whileTrue:req=self.hireader.gets()ifnotreq:breakres=process(req)self.transport.write(res)asyncdefmain():loop=asyncio.get_running_loop()server=awaitloop.create_server(lambda:RedisServerProtocol(),'0.0.0.0',5003)print('serving on {}'.format(server.sockets[0].getsockname()))asyncwithserver:awaitserver.serve_forever()try:importuvloopuvloop.install()exceptImportError:print('uvloop is not available')asyncio.run(main())Above code gives me:
redis-benchmark -p 5003 -t get -n 100000 -r 100000000...138696.25 requests per secondcompared to
redis-benchmark -p 6379 -t get -n 100000 -r 100000000...134589.50 requests per second
For further actions, you may consider blocking this person and/orreporting abuse
