Taking Django Distributed

Taking Django Distributed

A talk I gave at DjangoCon US 2017.

077e9a0cb34fa3eba2699240c9509717?s=128

Andrew Godwin

August 16, 2017
Tweet

Transcript

  1. Taking Django Distributed Andrew Godwin @andrewgodwin Taking Django Distributed

  2. Hi, I’m Andrew Godwin • Django core developer • Senior

    Software Engineer at • Needs to stop running towards code on fire
  3. Computers hate you.

  4. This makes distributed hard.

  5. 2001: A Space Odyssey Copyright Warner Brothers

  6. It’s time to split things up a bit. But how?

    And why?
  7. Code Databases Team

  8. There is no one solution

  9. Read-heavy? Write-heavy? Spiky? Predictable? Chatty?

  10. Code

  11. Use apps! They’re a good start! Ignore the way I

    wrote code for the first 5 years of Django.
  12. Formalise interfaces between apps Preferably in an RPC style

  13. Split along those interfaces Into separate processes, or machines

  14. Inventory Payments Presentation

  15. How do you communicate? HTTP? Channels? Smoke signals?

  16. None
  17. None
  18. None
  19. None
  20. Databases

  21. Users Vertically Partitioned Database Images Comments

  22. Main DB Replica Replica Replica Single main database with replication

  23. Partition Tolerant Available Consistent

  24. Non-consistency is everywhere It’s sneaky like that

  25. National Museum of American History

  26. Load Balancing

  27. Equally balanced servers Consistent load times Similar users

  28. Split logic Different processor loads Wildly varying users

  29. Reqs Time

  30. Reqs Time

  31. W E B S O C K E T S

  32. W E B S O C K E T S

    • They can last for hours • There’s not many tools that handle them • They have 4 different kinds of failure
  33. Design for failure, and then use it! Kill off sockets

    early and often.
  34. Team

  35. Developers are people too! They need time and interesting things

  36. Technical debt can be poisonous But you need a little

    bit to compete
  37. Single repo? Multiple repos? Each has distinct advantages.

  38. Teams per service? Split responsibility? Do you split ops/QA across

    teams too?
  39. Ownership gaps They’re very hard to see.

  40. Strategies

  41. Don’t go too micro on those services It’s easier in

    the short term, but will confuse you in the long term.
  42. Communicate over a service bus Preferably Channels, but you get

    to choose.
  43. Work out where to allow old data Build in deliberate

    caching or read only modes
  44. Design for future sharding Route everything through one model or

    set of functions
  45. Expect long-polls/sockets to die Design for load every time, and

    treat as a happy optimisation
  46. Independent, full-stack teams From ops to frontend, per major service

  47. Architect as a part-time position You need some, but not

    in an ivory tower
  48. 2001: A Space Odyssey Copyright Warner Brothers

  49. Maybe, just maybe, keep that monolith A well maintained and

    separated one beats bad distributed
  50. Thanks. Andrew Godwin @andrewgodwin aeracode.org