python进阶11并发之七多种并发方式的效率测试

原创博客地址：python进阶11并发之七多种并发方式的效率测试

测试map,apply_async,gevent协程爬虫
测试代码:网页爬虫

函数代码

 

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50

def thread_multi(): threads = list() for url in urls: threads.append(threading.Thread(target=process, args=(url,))) [t.start() for t in threads] [t.join() for t in threads] def thread_map(): pool = ThreadPool(max(1, cpu_count() - 1)) results = pool.map(process, urls) pool.close() pool.join() print(results) def thread_async(): pool = ThreadPool(max(1, cpu_count() - 1)) results = list() for url in urls: results.append(pool.apply_async(process, args=(url,))) pool.close() pool.join() print([result.get() for result in results]) def process_multi(): processes = list() for url in urls: processes.append(Process(target=process, args=(url,))) [t.start() for t in processes] [t.join() for t in processes] def process_map(): pool = Pool(processes=max(1, cpu_count() - 1)) results = pool.map(process, urls) pool.close() pool.join() print(results) def process_async(): pool = Pool(processes=max(1, cpu_count() - 1)) results = list() for url in urls: results.append(pool.apply_async(process, (url,))) pool.close() pool.join() print([result.get() for result in results])

测试结果concurrentOpt

 

 

	thread_multi	thread_map	thread_async	process_multi	process_map	process_async
0	5.732065	9.236784	7.831096	9.954077	9.778723	12.086315
1	4.868261	14.948347	8.431347	9.679722	17.086732	6.354689
2	7.17074	9.833528	12.248446	6.584711	17.405191	17.600024
3	11.223755	10.848167	9.235662	6.841372	9.969995	11.37249
4	10.391303	8.540373	8.236726	10.971645	8.964562	9.265784
5	8.3693	9.942565	9.541138	8.789822	8.266148	10.571744
6	8.799133	10.436757	13.669565	10.497021	9.668785	10.168379
7	5.603222	9.04568	9.843495	4.587275	14.596141	10.470989
8	9.003843	6.43141	9.941858	4.738146	8.170778	9.773284
9	9.414749	15.56822	9.23152	8.254023	8.781076	14.082026
10	8.0576371	10.4831831	9.8210853	8.0897814	11.2688131	11.1745724
最后一行为均值

测试结果concurrentOptGevent

 

 

	thread_multi	thread_map	thread_async	process_multi	process_map	process_async	gevent_test
0	10.770623	10.072167	11.220298	11.308327	3.40E-05	1.00E-05	6.035623
1	5.628367	8.850531	8.939288	7.608235	3.20E-05	1.00E-05	6.700398
2	5.214341	6.726455	13.69806	6.808565	3.20E-05	1.10E-05	6.868222
3	6.849362	7.976406	6.922554	5.7132	1.70E-05	5.00E-06	3.650169
4	5.442727	8.179533	8.46556	5.084351	3.00E-05	1.10E-05	7.655325
5	7.949327	9.9421	9.234288	4.723601	3.30E-05	1.00E-05	4.739602
6	5.765848	9.534865	7.956348	9.004707	2.00E-05	6.00E-06	5.40825
7	3.89613	10.026686	6.81114	5.685556	3.60E-05	1.00E-05	4.534598
8	8.838316	10.292949	6.709802	9.328648	1.80E-05	5.00E-06	5.724812
9	7.144312	9.321319	9.64821	5.898414	3.30E-05	1.00E-05	8.251311
10	6.7499353	9.0923011	8.9605548	7.1163604	2.85E-05	8.80E-06	5.956831
最后一行为均值

总结

concurrentOpt	进程or线程	同步or异步(不大确定)	阻塞or非阻塞(不大确定)	平均时间
thread_multi	多线程	异步	非阻塞	8.0576371
thread_map	线程池	(批)同步	阻塞	10.4831831
thread_async	线程池	异步	非阻塞	9.8210853
process_multi	多进程	异步	非阻塞	8.0897814
process_map	进程池	（批）同步	阻塞	11.2688131
process_async	进程池	异步	非阻塞	11.1745724

concurrentOptGevent	进程or线程	同步or异步(不大确定)	阻塞or非阻塞(不大确定)	平均时间
thread_multi	多线程	异步	非阻塞	6.7499353
thread_map	线程池	（批）同步	非阻塞	9.0923011
thread_async	线程池	异步	非阻塞	8.9605548
process_multi	多进程	异步	非阻塞	7.1163604
process_map	进程池	（批）同步	非阻塞	卡住
process_async	进程池	异步	非阻塞	卡住
gevent_test	协程	异步	非阻塞	5.956831

结论:
01，启用gevent后，除了卡住的，线程和进程均加快1s左右时间
02，协程在线程程序中是最快的
03，多线程程序下载速度弱优于多进程
04，不论是进程还是线程，使用thread_async都快于map
05，不考虑协程时，多线程较线程池速度更快，多进程较进程池速度更快，这一点不大符合理论，个人感觉和url数量少有关.

至于进程池在启用gevent后卡住的问题，网上也没查到相关的靠谱资料，哪位大牛晓得的话，求解释～
测试代码:github的concurrentOpt.py和concurrentOptGevent.py