Scaling django
Download as PPTX, PDF0 likes80 views
This document discusses how to scale Django applications. It covers concurrency models like multi-process, threads, and event-driven approaches. It then focuses on using gevent, an asynchronous networking library, to add concurrency to Django. Gevent uses greenlets (lightweight coroutines) and event loops and allows blocking code to run concurrently. It also discusses deploying Django applications with Gunicorn and gevent to improve scalability over alternatives like Twisted.
![Coroutines
Python coroutines are almost similar to
generators.
def abc(seq):
lst = list(seq)
for i in lst:
value = yield i
if cmd is not None:
lst.append(value)
r = abc([1,2,3])
r.send(4)](https://image.slidesharecdn.com/scalingdjango-170714062328/85/Scaling-django-9-320.jpg)
![Gevent code
import gevent
from gevent import socket
urls = [‘www.google.com’, ‘www.python.org’]
jobs = [gevent.spawn(socket.gethostbyname, url) for url in urls]
gevent.joinall(jobs, timeout=2)
[job.value for job in jobs]
Output: ['172.217.26.68', '151.101.192.223']](https://image.slidesharecdn.com/scalingdjango-170714062328/85/Scaling-django-16-320.jpg)
![Migrating our products
Twisted Code:
def handle_purchase( request ):
defs = []
defs.append(billert())
defs.append(ads())
defs.append(recos() )
defs.append(meta() )
def = DeferredList(defs, …)
Return NOT_DONE_YET
Gevent Code:
def handle_purchase( request ):
jobs = []
jobs.append( gevent.spawn( biller, ... ) )
jobs.append( gevent.spawn( ads, ... ))
jobs.append( gevent.spawn( meta, ... ) )
jobs.append( gevent.spawn( reco, ... ) )
gevent.joinall()](https://image.slidesharecdn.com/scalingdjango-170714062328/85/Scaling-django-23-320.jpg)
Scaling django
- 1. How to scale Django Ahsanuzzaman Khan Sr. Software Engineer Z. S. Solution Ltd.
- 2. How does Django application scale?
- 4. Models Found around the web! Multi-process Threads Event driven Coroutines
- 5. Process/Threads There are blocking sections in the code 1. Python Global Interpreter Locator (GIL) is an issue in thread based concurrency request dispatch worker_1 worker_n sock_wr() db_rd_wr() read(fp)
- 6. Event Driven Can anyone gives an example? event_1 event_n block_on_events( ) h_n() h_2() h_1()
- 7. Event Driven Web Servers 1. Nginx 2. Tornado 3. Node.js event_1 event_n block_on_events( ) h_n() h_2() h_1()
- 8. Python Twisted 1. An event driven library 2. Using epoll or kqueue client_1 client_n Nginx (SSL & LB) s_n() s_2() s_1()
- 9. Coroutines Python coroutines are almost similar to generators. def abc(seq): lst = list(seq) for i in lst: value = yield i if cmd is not None: lst.append(value) r = abc([1,2,3]) r.send(4)
- 10. Gevent A coroutine-based Python networking library that uses greenlet to provide a high-level synchronous API on top of the libevent event loop.
- 12. Gevent Features 1. Fast event-loop based on libevent (epoll, kqueue etc.) 2. Lightweight execution units based on greenlets (coroutines) 3. Monkey patching support 4. Simple API 5. TCP/UDP/HTTP and Fast WSGI servers 6. Cooperative socket with SSL support
- 13. Greenlets 1. Micro-threads with no implicit scheduling 2. It is nothing just coroutines 3. Other systems like gevent build micro- threads on top of greenlets 4. Execution happens by switching execution among greenlet stacks but not implicit Main greenlet task() Child greenlet h_n() pause() func() res()
- 14. Greenlet code example from greenlet import greenlet def test1(): print 12 gr2.switch() print 34 def test2(): print 56 gr1.switch() print 78 gr1 = greenlet(test1) gr2 = greenlet(test2) gr1.switch() Output: 12 56 34
- 15. How does gevent work 1. Creates an implicit event loop inside a dedicated greenlet 2. When a function in gevent wants to block, it switches to the greenlet of the event loop. This will schedule another child greenlet to run. 3. The eventloop automatically picks up the fastest polling mechanism available in the system. 4. One event loop runs inside a single OS thread (process).
- 16. Gevent code import gevent from gevent import socket urls = [‘www.google.com’, ‘www.python.org’] jobs = [gevent.spawn(socket.gethostbyname, url) for url in urls] gevent.joinall(jobs, timeout=2) [job.value for job in jobs] Output: ['172.217.26.68', '151.101.192.223']
- 17. Gevent apis 1. Greenlet management (spawn, timeout, schedule) 2. Greenlet local data 3. Networking (socket, ssl, dns, select) 4. Synchronization a. Event - notify multiple listeners b. Queue - synchronized producer/consumer queues c. Locking - Semaphores 5. Greenlet pools 6. TCP/IP and WSGI servers and many more...
- 18. Gevent advantages 1. Almost synchronous code. No callbacks and deferred. 2. Lightweight greenlets. 3. Good concurrency. 4. No issues of python GIL. 5. No need for in-process locking, since a greenlet cannot be preempted.
- 19. Gevent issues 1. A greenlet will run till it blocks or switches a. May vary of large/infinite loops 2. Monkey patching is required for un-supported blocking libraries. Might not work well with some libraries.
- 20. The Django Problem 1. Asynchronous not yet supported 2. In HTTP request cycle we want following operations a. Fetch some metadata for an item being sold b. Purchase the item for the user in the billing system c. Fetch ads to be shown along with the item d. Fetch recommendations based on this item 3. In parallel … !! a. Twisted was the only option Gevent saved the day :)
- 21. Twisted Issues The issues were with everything else Header management Templates for response ORM support Session management and auth Caching support Many more…. The above are django’s strength Django has vibrant eco-system
- 22. Gunicorn 1. Python WSGI HTTP server 2. Supports running code under worker, eventlet, gevent etc. a. Uses monkey patching 3. Excellent django support 4. Enabled gevent support for our app by default without any code changes 5. Spawns and manages worker processes and distributes load among
- 23. Migrating our products Twisted Code: def handle_purchase( request ): defs = [] defs.append(billert()) defs.append(ads()) defs.append(recos() ) defs.append(meta() ) def = DeferredList(defs, …) Return NOT_DONE_YET Gevent Code: def handle_purchase( request ): jobs = [] jobs.append( gevent.spawn( biller, ... ) ) jobs.append( gevent.spawn( ads, ... )) jobs.append( gevent.spawn( meta, ... ) ) jobs.append( gevent.spawn( reco, ... ) ) gevent.joinall()
- 24. Migrating our products 1. Time is concern :(
- 25. Deployment Client Nginx (SSL & LB) Gunicorn 1 Gunicorn N Proc N Proc 2 Proc 1
- 26. Conclusion 1.Single framework for all products 2.Use django’s awesome features and ecosystem 3.Increased scalability. More so with celery. 4.Use blocking python libraries without worrying too much 5.No more usage of python-twisted 6.Coding, testing and maintenance is much easier