kuafu
kuafu

Reputation: 1486

Why does gevent use more memory than threading

I think gevent should use less memory than threading, but actually it cost more memory than threading.

here is my code: the gevent

#import gevent.monkey
#import gevent.httplib as ghttplib
import httplib as ghttplib
import httpsqs
#gevent.monkey.patch_all()
#from urlparse import urlparse
#from gevent.pool import Pool
#import gevent
#import MySQLdb
import urllib2
#from MySQLdb.cursors import SSCursor
#import gevent_profiler
import requests
import time
from threading import Thread
#import Queue
import os
import memory
import sys
#gevent_profiler.print_percentages(True)
#gevent_profiler.time_blocking(True)
#gevent_profiler.set_stats_output('my-stats.txt')

user_agent = 'Mozilla/5.0 (Windows NT 6.1; rv:10.0)\
        Gecko/20100101 Firefox/10.0'
headers = { 'User-Agent' : user_agent }


scale = [1,5,10,20,50,100,200,300]

data = open("thread.txt",'w')
#db=MySQLdb.connect(host='125.221.225.12',user='root',passwd='young001',charset='utf8',db='delicious',use_unicode=True) 
#cur = db.cursor()
print os.getpid()

def get(url):
    r = requests.get(url,headers=headers,timeout=10)
    return r

if(os.path.exists("./urls_httpsqs")):
    pass
else:
    os.makedirs("./urls_httpsqs")


class URLThread(Thread):
    def __init__(self, queue, queue_name, timeout=10, allow_redirects=True):
        super(URLThread, self).__init__()
        #self.url = url
        self.timeout = timeout
        self.runflag = True
        self.allow_redirects = allow_redirects
        self.response = None
        self.headers = { 'User-Agent' : user_agent }
        #self.db = MySQLdb.connect(host='125.221.225.12',user='root',passwd='young001',charset='utf8',db='delicious',use_unicode=True) 
        #self.cur = self.db.cursor()
        self.queue_name = queue_name
        self.queue = queue

    def save_disk(self,res,pid):
        datafile = open("./urls_httpsqs/%s"%pid,"w")
        datafile.write(res.content)
        datafile.close()

    def run(self):
        while self.runflag:
            url = self.queue.get(self.queue_name).strip()
            if httpsqs.isOK(url):
                pass
            else:
                return
            #print "getting",url
            try:
                self.response = requests.get(url, timeout = self.timeout, headers = self.headers, allow_redirects = self.allow_redirects)
                #pid = url.split("/")[-1]
                #print "pid is", pid
                #self.save_disk(self.response,pid)
                #print "file done"

            except Exception , what:
                print what
                #self.insert_into_fail(db,url)
                pass
            #finally:
                #self.queue.task_done()
    def stop(self):
        self.runflag = False

#queue = Queue.Queue(50)
queue = httpsqs.Httpsqs("125.221.225.12")
queue_name = "coroutine"


#gevent_profiler.attach()
threads = []

now = time.time()

for num in scale:
    for i in range(num):
        threads.append(URLThread(queue,queue_name))
    #for t in threads:
        #t.start()
    for t in threads:
        t.stop()
    print memory.resident()
    threads = []
    data.write(str((memory.resident()/1000000)))
    data.write("\t")
    data.write(str((memory.resident()/1000000)+memory.memory()/1000000))
    data.write("\n")
    data.flush()

#sys.exit(0)

#for t in threads:
    #t.join()

end = time.time()
print "virtual memory is", memory.memory()
print "resident memory is", memory.resident()
print "stack memory is", memory.stacksize()
print "begin is",now
print "end is",end
print "it costs", end-now

the threading:

import gevent.httplib as ghttplib
import time
import httplib as ghttplib
import httpsqs
#gevent.monkey.patch_all()
#from urlparse import urlparse
#from gevent.pool import Pool
#import gevent
import MySQLdb
import urllib2
#from MySQLdb.cursors import SSCursor
#import gevent_profiler
import requests
from threading import Thread
import multiprocessing
#import Queue
import os
import memory

#gevent_profiler.print_percentages(True)
#gevent_profiler.time_blocking(True)
#gevent_profiler.set_stats_output('my-stats.txt')

user_agent = 'Mozilla/5.0 (Windows NT 6.1; rv:10.0)\
        Gecko/20100101 Firefox/10.0'
headers = { 'User-Agent' : user_agent }

scale = [1,5,10,20,50,100,200,300]
data = open("process.txt",'w')

#db=MySQLdb.connect(host='125.221.225.12',user='root',passwd='young001',charset='utf8',db='delicious',use_unicode=True) 
#cur = db.cursor()
total_mem = 0

def get(url):
    r = requests.get(url,headers=headers,timeout=10)
    return r

if(os.path.exists("./urls_httpsqs")):
    pass
else:
    os.makedirs("./urls_httpsqs")

def save_disk(res,pid):
    datafile = open("./urls_httpsqs/%s"%pid,"w")
    datafile.write(res.content)
    datafile.close()

def run(queue,queue_name):
    #print os.getpid()

    #print 'total mem is', total_mem
    while True:
        url = queue.get(queue_name).strip()
        if httpsqs.isOK(url):
            pass
        else:
            return

        #print "getting",url
        try:
            #db = MySQLdb.connect(host='125.221.225.12',user='root',passwd='young001',charset='utf8',db='delicious',use_unicode=True) 
            #response = requests.get(url, timeout = 10)
            response = requests.get(url)
            #pid = self.insert_into_avail(db,url)
            #pid = url.split("/")[-1]
            #save_disk(response,1)

        except Exception , what:
            print what
            #self.insert_into_fail(db,url)
            pass



queue = httpsqs.Httpsqs("125.221.225.12")
queue_name = "coroutine"
#print os.getpid()

#gevent_profiler.attach()
now = time.time()
record = []
for num in scale:
    for i in range(num):
        process = multiprocessing.Process(target=run,args=(queue,queue_name))
        process.start()
        record.append(process)
    for i in record:
        i.terminate()
    record = []
    print "done"
    print memory.resident()
    print num
    print memory.resident()*num
    data.write(str((memory.resident()*num/1000000)))
    data.write("\t")
    data.write(str((memory.resident()*num/1000000)+memory.memory()/1000000))
    data.write("\n")
    data.flush()
    #for process in record:
        #process.join()
#pool.close()
#pool.join()
data.close()
end = time.time()

print "virtual memory is", memory.memory()
print "resident memory is", memory.resident()
print "stack memory is", memory.stacksize()
print "begin is",now
print "end is",end
print "it costs", end-now

and I use the following to know the memory cost:

import os
import sys

sys_pid = sys.argv[1]
sys_pid = int(sys_pid)
#_proc_status = '/proc/%d/status' % os.getpid()
_proc_status = '/proc/%d/status' %sys_pid 

_scale = {'kB': 1024.0, 'mB': 1024.0*1024.0,
          'KB': 1024.0, 'MB': 1024.0*1024.0}

def _VmB(VmKey):
    '''Private.
    '''
    global _proc_status, _scale
     # get pseudo file  /proc/<pid>/status
    try:
        t = open(_proc_status)
        v = t.read()
        t.close()
    except:
        return 0.0  # non-Linux?
     # get VmKey line e.g. 'VmRSS:  9999  kB\n ...'
    i = v.index(VmKey)
    v = v[i:].split(None, 3)  # whitespace
    if len(v) < 3:
        return 0.0  # invalid format?
     # convert Vm value to bytes
    return float(v[1]) * _scale[v[2]]


def memory(since=0.0):
    '''Return memory usage in bytes.
    '''
    return _VmB('VmSize:') - since


def resident(since=0.0):
    '''Return resident memory usage in bytes.
    '''
    return _VmB('VmRSS:') - since


def stacksize(since=0.0):
    '''Return stack size in bytes.
    '''
    return _VmB('VmStk:') - since

print "virtual memory is", memory()
print "resident memory is", resident()
print "stack memory is", stacksize()

this come from the python cookbook.

and the output is: thread

8.310784    23.42912
8.347648    23.445504
8.35584 23.457792
8.368128    23.47008
8.41728 23.519232
8.503296    23.601152
8.671232    24.117248
8.843264    24.293376

gevent

9.019392    24.829952
9.048064    24.846336
9.056256    24.854528
9.07264 25.14944
9.1136  25.1904
9.19552 25.27232
9.330688    25.407488
9.46176 25.92768

where did I do wrong?

Upvotes: 4

Views: 1417

Answers (1)

navendu
navendu

Reputation: 274

Gevent is bound to use more memory as it maintains its own light weight threads (greenlets) which is bound to cause some overhead.

If your application is CPU bound, then gevent might not be that useful for you.

But if your application is I/O bound, then gevent is awesome as you can reach concurrency levels of 1000s on 4-8 GiG machine.

Also as my friend once said, memory is expensive but not that expensive :-) Cheers!

Upvotes: 1

Related Questions