Tornado cherrypy bottle performance test

Tornado cherrypy bottle performance test

2011-11-25 I made a comparison between tornado cherrypy and bottle today ab -n 10000 -c 1000 http://192.168.8.2:8889/1588req/s (tornado) ab -n 10000 -c 1000 http://192.168 .8.2:9080/hello/shengp 323.35 req/s (bottle) ab -n 10000 -c 1000 http://192.168.8.2:8088/nqsdk/change 452.36 req/s (cherrypy) I found out that I tested 3000+ with others A big gap. The posts posted in CPyUG were included in the Woodpecker community. Very excited. Link: http://wiki.woodpecker.org.cn/moin/MiscItems/2011-11-25 Dean forever dean forever deanforwever2010@gmail .com sender's local time: sending time 10:47 (GMT+08:00). The current time of the sending place: 10:05 PM. ✆ Subject: [CPyUG] When testing the performance of tornado and adding it to 2000, it starts to fail? Dual-core 2.0Ghz debian system runs tornado's hello world: switch line number display import tornado.httpserver import tornado.ioloop import tornado.web class MainHandler(tornado.web.RequestHandler): def get(self): self.write("Hello, world") application = tornado.web.Application([(r"/", MainHandler),]) if __name__ == "__main__": web from tornado import httpserver from tornado import ioloop from tornado.web import RequestHandler class non(RequestHandler): def get(self): """docstring for get""" self.write("hello") def main(): " ""docstring for main""" sokets = tornado.netutil.bind_sockets(8080) tornado.process.fork_processes(0) application = tornado.web.Application([ ('',non)]) http_server = httpserver.HTTPServer(application ) http_server.add_sockets(sokets) ioloop.IOLoop.instance().start() if __name__ =='__main__': main() You are the best! ab -n 10000 -c 1000 http://192.168.0.36:1081/... Requests per second: 8416.98 [#/sec] (mean) If you open concurrently to 5000, you can still reach 5000req/s. It seems to be for my own use. Wrong, hehe Feilong analysis Felinx Lee felinx.lee@gmail.com Sender's local time: Sending time 11:33 (GMT+08:00). The current time of the sending place: 10:PM 10. ✆ This is controlled by the fd limit allowed by the system, you can refer to this http://thedaneshproject.com/posts/how-to-increase-total-file-descriptors-count-on-linux/The fd limit of each process has Restrictions, refer to: http://stackoverflow.com/questions/2569620/socket-accept-error-24-to-many-open-files using fork is actually based on multi-core to run multiple processes, each process is limited It has not been reached yet, so there is no error. But the bosses of tornado do not recommend using fork. In the end, according to WooParadog's method, the best concurrency reached 1W+ (on an 8-core system). I followed Feilong's recommended article and found that the effect is not working. I don't know if it is a machine problem. . . There is a problem with the 404 error. http://192.168.0.36:1081 Remove the slash and run without error: Requests per second: 11187.92 [#/sec] (mean)

Reference: https://cloud.tencent.com/developer/article/1067258 tornado cherrypy bottle performance test-Cloud + Community-Tencent Cloud