My multi-threading code with django connection pool doesn't have any improvement

Question

I am struggling with multi-threading with a connection pool on Django.

I know python threading has GIL issue but I thought python threading is enough to improve performance if the most of the work are DB I/O.

First all I tried to implement a small code to prove my thought.

Simply explaining, the code uses threadPool.apply_async() with a DB connection pool set by CONN_MAX_AGE in settings.py.

With the code, I repeat controlling the number of threads for the worker thread.

from multiprocessing            import pool
from threadPoolTestWithDB_IO    import models
from django.db                  import transaction
import django
import datetime
import logging
import g2sType


def addEgm(pre, id_):
    """
    @summary: This function only inserts a bundle of records tied by a foreign key 
    """
    try:
        with transaction.atomic():

            egmId = pre + "_" + str(id_)
            egm = models.G2sEgm(egmId=egmId, egmLocation="localhost")
            egm.save()

            device = models.Device(egm=egm,
                          deviceId=1,
                          deviceClass=g2sType.t_deviceClass.G2S_eventHandler,
                          deviceActive=True)
            device.save()

            models.EventHandlerProfile(device=device, queueBehavior="a").save()
            models.EventHandlerStatus(device=device).save()

            for i2 in range(1, 200):
                models.EventReportData(device=device,
                                       deviceClass=g2sType.t_deviceClass.G2S_communications,
                                       deviceId=1,
                                       eventCode="TEST",
                                       eventText="",
                                       eventId=i2,
                                       transactionId=0
                                       ).save()

            print "Done %d" % id_


    except Exception as e:
        logging.root.exception(e)


if __name__ == "__main__":

    django.setup()
    logging.basicConfig()

    print "Start test"

    tPool = pool.ThreadPool(processes=1)    #Set the number of processes

    s = datetime.datetime.now()
    for i in range(100):                    #Set the number of record bundles
        tPool.apply_async(func=addEgm, args=("a", i))

    print "Wait worker processes"
    tPool.close()                           
    tPool.join()

    e = datetime.datetime.now()
    print "End test"

    print "Time Measurement : %s" % (e-s,)

    models.G2sEgm.objects.all().delete()    #remove all records inserted while the test
--------------------------
# settings.py


DATABASES = {
             'default': {
                         'ENGINE': 'django.db.backends.oracle',
                         'NAME': 'orcl',
                         'USER': 'test',
                         'PASSWORD': '1123',
                         'HOST': '192.168.0.90',
                         'PORT': '1521',
                         'CONN_MAX_AGE': 100,
                         'OPTIONS': {'threaded': True}
                         }
             }

However, the result came out as they don't have any big difference between 1 thread worker and multi-thread works.

For example, It takes 30.6 sec with 10 threads and takes 30.4 sec with 1 thread.

What did I go wrong?

Answer 1

Either you have problems on database level. You can prove it by execution this query:

select /* +rule */
    s1.username || '@' || s1.machine
    || ' ( SID=' || s1.sid || ' ' || s1.program || ' )  is blocking ' || s2.username || '@' || s2.machine || ' ( SID=' || s2.sid || ' ' || s2.program || ' ) ' AS blocking_status
    from v$lock l1, v$session s1, v$lock l2, v$session s2
    where s1.sid=l1.sid and s2.sid=l2.sid
    and l1.BLOCK=1 and l2.request > 0
    and l1.id1 = l2.id1
    and l2.id2 = l2.id2 ;

Or there are threads being blocked in Python. (possibly on DB driver level). Attach gdb to python process and then execute thread apply all bt.

And you will see.