为什么multiprocess .join（）会减慢速度？

Question

我正在从API端点获取数据，这是我应用程序的瓶颈。

我的问题是： 为什么添加.join（）会大大降低程序速度？ 我知道，为了确保主节点在子进程“正常”退出后退出，是安全的。我只是不明白为什么它比运行此单线程的速度还要慢。

在我要馈入多进程队列的文件中，包含n行文件名，然后将其提取到API。 然后将该请求的输出插入到另一个队列，该队列在mp.queue进程准备就绪（写入文件）之后运行。

from multiprocessing import Process, Queue
import time
import sys
import requests
import json
import xlrd
import openpyxl



def reader_proc (queue):
    ## Read from the queue; this will be spawned as a separate Process
    while True:
        msg = queue.get()  # Read from the queue and do nothing
        if (msg == 'DONE'):
            break
        r = search('filename:' + msg[1], token)
        try:
            if r.json()['totalHits'] > 0:

                ad = r.json()['hits'][0]
                entryQueue.put((msg[0], msg[1], 'x',
                                ad.get('metadata', {}).get('cf_eanCode', ''),
                                ad.get('metadata', {}).get('cf_assetPriority', ''),
                                ad.get('id', ''),
                                ad.get('metadata', {}).get('assetPath', ''),
                                r.json()['totalHits']
                                ))
            else:
                entryQueue.put((msg[0], msg[1]))

        except KeyError:
            pass


def writer (filename, queue):

    queue.put(filename)  # Write to queue
    queue.put('DONE')



missing_xcodes = '/missing_xcodes.txt'

xlxsf = 'xeancodes_mp.xlsx'
xfile = openpyxl.load_workbook(xlxsf)
sheet = xfile.get_sheet_by_name('xcodes')

# Multiprocess will throw data to this queue
entryQueue = Queue()

if __name__ == '__main__':
    _start = time.time()
    pqueue = Queue()  # Processing queue for MP

    with open(missing_xcodes) as f:
        for i, filename in enumerate(f):
            filename = filename.replace("\n", "")
            filename = filename.strip()
            reader_p = Process(target=reader_proc, args=((pqueue),))
            reader_p.daemon = True
            reader_p.start()
            # reader_p.join() # I've kept reader here SUPER SLOW.
            writer((i, filename), pqueue)  # Send to queue.

       # reader_p.join() # Tried here, too. Will skip some of the queue 
       elements that are not ready, so its not working as should..

    time.sleep(2) # Need to use sleep here to make sure all child processes are 
    # done.
    entryQueue.put('DONE') # This will add STOP to the queue as a last in FIFO 
    # queue.

    while True:
        msg = entryQueue.get()  # Read from the queue
        if (msg == 'DONE'):
            break

        nro = msg[0] + 1 # In excel things start @1
        sheet['A' + str(nro)] = str(msg)  # filename

    xfile.save(xlxsf)
    print(time.time() - _start)