Queue Everything and Please Everyone

The intent of this talk is to introduce the concepts of queueing in web development, tools at one’s disposal, some common patterns and how it all can be used to reduce time to generate user feedback.

Transcript

1. Queue Everything & Please Everyone * * * Vaidik Kapoor
2. Who am I? ➔ Undergraduate student. ➔ Involved with: ✔ Mozilla as a Contributor and Rep ✔ Drupal as a Contributor ➔ Twitter: @vaidikkapoor ➔ Github: vaidikkp ➔ Web: vaidikkapoor.info
3. Web Applications
4. Responsiveness
5. Why do everything at once? Task 1 Task 2 Task 3 Task N Client Request Response
6. Break the Request-Response Cycle Task 1 Task 2 Task 3 Task N Client Request Response
7. Meet Queues
8. What can Queues help with? ➔ Background Processing ✔ Data Processing ✔ Media Processing ✔ Updating Caches ➔ Anything that you want to offload off your server. ➔ Improve the overall User Experience
9. Redis + HotQueue / PyRes
10. Redis + HotQueue PRODUCER: from hotqueue import HotQueue queue = HotQueue("myqueue", host="localhost", port=6379, db=0) queue.put(message) CONSUMER / WORKER: from hotqueue import HotQueue queue = HotQueue("myqueue", host="localhost", port=6379, db=0) while True: message = queue.get() # do something awesome
11. Redis + PyRes ➔ Python clone of Github's Resque ➔ Offers a lot more features than HotQueue ✔ Job failure handling ✔ Queue status information ➔ Comes with Monitoring System written in Ruby
12. Redis + PyRes PRODUCER: from pyres import ResQ r = Resq() class WebPage: queue = “screenies” @staticmethod def perform(urll): # save the screenshot r.enqueue(WebPage, 'http://python.org') CONSUMER: ./pyres_worker screenies
13. Message Queue A system for enabling asynchronous processing of discrete tasks.
14. What can MQs help with? ➔ Everything that generic Queues can help with. ➔ Increase Reliability ➔ Cron Jobs (Celery)
15. Available MQ Solutions ➔ RabbitMQ ➔ Amazon Simple Queue ➔ Apache MQ ➔ Gearman ➔ Starling ➔ OpenAMQ ➔ Sun Java Message Queue System
16. Message Queue Protocols ➔ AMQP: Advanced Message Queue Protocol ➔ JMS: Java Messaging Service ➔ STOMP: Streaming Text Oriented Messaging Protocol
17. Criteria for Broker Selection ➔ Difficulty in Recovery ➔ Relatively low level of required maintenance ➔ Ease of Deployment ➔ Durability ➔ Persistence ➔ Community Support ➔ Cluster Support ➔ What language is it written in?
18. RabbitMQ ➔ A Message Broker ➔ Implements AMQP
19. Pika & Other AMQP Libraries
20. Celery ➔ An amazing Task Manager ➔ Batteries Included ➔ Uses RabbitMQ as broker (or almost anything) ➔ Libraries for most of the common web frameworks like Django, Flask, Pyramid ➔ Supports multiprocessing. ➔ Distribute tasks easily. ➔ Makes life easy
21. Celery Example from celery.task import task @task(): def take_screenshot(url): # magical code to take screenshot comes here
22. How to put it all together?
23. Web App Broker Task Manager Worker Server (s) Database The General Setup
24. Then what?
25. Notify Your Users
26. Django's Messaging System
27. Flask's Flashes
28. Poll using AJAX
29. WebSockets
30. Email
31. Things to Remember
32. Isolate Isolate code, make reusable components.
33. Recycle Remove Request Dependency
34. Unit Testing?
35. Review ➔ Choose a system according to needs ➔ Build a robust system ➔ Integrate seamlessly with your UI ➔ Isolate & Recycle: make it a habit ➔ Improve the overall UX! ➔ Please everyone, even yourself!
36. References ➔ PyRes - http://pyres.readthedocs.org/ ➔ RabbitMQ - http://www.rabbitmq.com/ ➔ Celery - http://celeryproject.org/ ➔ Amazing Article - http://wiki.secondlife.com/wiki/Message_Queue_Evaluation_Notes
37. Thank You!