Sending SMS and email to 100000 members in django-mysql envrionment - django

Question

I am having a real headache on how to handle this one. Basically, the application got members that are projected to reach one million at the end of the year. It relies heavily on USSD but also have email. Actually for now, I would prefer to send the SMS first.
The issue is this: the members have groups based on their activities and a single member can have multiple groups. Currently, the highest number of members in a group is 17,000.00. The group can basically send SMS to those 17,000 members. The group leaders specify paramters ("All Members","Females","Age 24-28" etc) and send the SMS, which must save a copy in the database. Currently, there are 5 active groups but they will certainly increase in the future and they can all request to broadcast SMS to members at once.
The phone numbers of members is kept in:
class Profile(model.Models):
user=models.ForeignKey(User)
phone=models.CharField(max_length=13)
Similarly, the app should basically scan the member profiles to send them period notifications. For now, I am following the following:
Select the phone numbers of all members that satisfy the criteria
Create an id for the broadcast and wait for previously stacked SMS requests to finish. Then add the selected phones to a secondary table referencing to the broadcast
Loop through each phone and send one by one. Once finished, mark the broadcast as finished
class BroadCast(models.Model):
code=models.CharField(max_length=50) #rand generated
group=models.ForeignKey(Corporate)
finished=models.IntegerField(default=0)
message=models.CharField(max_length=200)
class Phone(models.Model):
broadcast=models.ForeignKey(BroadCast)
But am disappointed by its performance especially for multiple requests. What can i do to improve it?
I am using twilio paid SMS.
tele=models.CharField(max_length=13)

Answer 1

I had an sms application before that had scaled and slowed down when the user base was large, to fix that we used queue for the sms sending task. Worked wonders and should also work perfectly in your case.