SE4FP - Elixir in Production - 2026

Transcript

1. beyond the code ELIXIR IN PRODUCTION SE4FP 2026
2. ELAINE NAOMI WATANABE Software Engineer/Developer (since 2008) Staff Software Engineer (since 2023) B.Sc. in Computer Engineering M.Sc. in Computer Science speakerdeck.com/elainenaomi Photo edited by AI :) linkedin.com/in/elainenaomi slides available on
3. This IS NOT a presentation about the adoption of the Elixir functional programming language.
4. It's about the CHALLENGES of developing and evolving Elixir systems in production FOR YEARS And I believe this applies to other languages as well
5. ELIXIR IN PRODUCTION
6. Phoenix — web and API Users Ecto PostgreSQL Basic Web Application E PL M A EX Elixir Application
7. Observability Telemetry Grafana Elixir Application Phoenix — web and API Users Observability Ecto More dependencies Oban — asynchronous jobs Data PostgreSQL Redis
8. Users Load Balancer Kafka Observability More dependencies More external services Greater scalability Higher data volume Web deployment Deployment worker Web 1 Worker 1 Worker 2 Ecto Ecto Phoenix Ecto Web 2 Phoenix Ecto Telemetry Grafana Datadog Oban Oban Broadway Broadway External APIs Stripe, S3, SendGrid BEAM cluster Shared data PostgreSQL Redis
9. Users Load Balancer Kafka Observabilidade Web deployment Deployment worker Web 1 Web 2 Worker 1 Worker 2 Phoenix Phoenix Ecto Ecto Ecto Ecto Telemetry Grafana Datadog Oban Oban Broadway Broadway APIs externas Stripe, S3, SendGrid BEAM cluster Shared data PostgreSQL Redis More developers
10. Users Load Balancer Kafka Observability Web deployment Deployment worker Web 1 Web 2 Worker 1 Worker 2 Phoenix Phoenix Ecto Ecto Ecto Ecto Telemetry Grafana Datadog Oban Oban Broadway Broadway External APIs Stripe, S3, SendGrid BEAM cluster Shared data PostgreSQL Redis And AI Agents
11. Users Load Balancer Kafka Observability Deployment web Deployment worker Web 1 Web 2 Worker 1 Worker 2 Phoenix Phoenix Ecto Ecto Ecto Ecto Telemetry Grafana Datadog Oban Oban Broadway Broadway External APIs Stripe, S3, SendGrid BEAM cluster Shared data PostgreSQL Redis Topic for another presentation :) And AI Agents
12. CHALLENGE #1: CODE
13. Learning the Elixir language and the functional paradigm
14. Learning the Elixir language and the functional paradigm
15. ON TI TA AP AD Language documentation, examples, prior knowledge of other stacks, etc., help mitigate this problem. People who already program in other languages.
16. And what happens after the adaptation period?
17. How does the code organization change over time?
18. Is the code still easy to evolve?
20. Functional programming does not prevent spaghetti code Code that is difficult to understand and evolve
21. A lack of familiarity with Elixir/OTP and with patterns can unnecessarily increase code complexity and lead to anti-patterns
22. https://sol.sbc.org.br/index.php/cbsoft_estendido/article/view/36959
23. https://elixir.hexdocs.pm/what-anti-patterns.html
24. I bet there are more patterns and anti-patterns to document :)
25. However…
26. A system isn't just about code!
27. CHALLENGE #2: PEOPLE
28. To keep existing systems in Elixir running in the future…
29. We need more developers!
30. We need more developers! and probably AI agents… but that's a topic for another presentation :)
31. To hire or to train people?
32. Elixir experts or developers from other languages? $$$
33. Yes, you can hire developers who know NOTHING about Elixir or functional programming but who already program in other languages
34. Other languages Elixir knowledge transfer
36. How to go beyond the basics in Elixir, OTP, and functional programming?
37. How do we train Elixir specialists? courses, events, books, etc.
38. How do we train Elixir specialists? We have good options, but there are still few compared to other languages
39. How do we train Elixir specialists? Who should invest in this? Companies, universities, developers?
40. What about those who are just starting to program?
41. What about those who are just starting to program? How can we train juniors if the content is only for professionals?
42. What about those who are just starting to program? How can we build teams with developers at different levels: junior, mid-level, senior?
43. |>
44. The challenges go beyond code and people
45. CHALLENGE #3: ECOSYSTEM
46. Everything around a language that helps us develop and maintain systems Ecosystem
47. Language + Libraries + Tools + Infrastructure + Community Ecosystem
49. But what if the problem isn't in your code?
50. When a dependency (library) is no longer maintained, do we replace it with another?
51. Do we have many options for that?
52. hex.pm
53. ~18.2 thousand packages
54. No updates ≠ abandoned but…
55. 54.2% of Mix packages in hex.pm DID NOT receive a new version in the last 3 years ~9.9 thousand packages Exploratory analysis based on packages with build_tool: mix available in Hex on 02/09/2026.
56. What if your system needs a change from a specific library? What to do in this case? Fork?
57. What if your system needs a change to a specific library? And what about the additional time/cost to maintain a fork?
58. JavaScript (npm) 2.7 million Python (PyPI) 885 thousand The gap becomes even Java (Maven Central) 808.6 thousand more apparent when compared to other Ruby (RubyGems) 196.7 thousand languages Elixir (Hex / Mix) 18.2 thousand
59. JavaScript (npm) 2.7 million Python (PyPI) 885 thousand Java (Maven Central) 808.6 thousand Ruby (RubyGems) 196.7 thousand Elixir (Hex / Mix) 18.2 thousand 10x more packages than Hex.pm (Mix)
60. Always monitor your dependencies And keep your system up to date
61. |>
62. Infrastructure
63. Elixir in production also means BEAM in production BEAM =~ Elixir/Erlang VM
64. Elixir / OTP Cloud Services Horizontal/vertical scalability Databases and messaging Observability / Instrumentation
65. Overall, Elixir runs well in production BEAM VM: 30+ years of evolution and use in production
67. But what happens when it doesn't behave as well in production?
68. How do you know what the expected behavior is in production? For example: memory consumption, CPU usage, etc.
69. How can you tell if it's a problem with the company's internal system or a common problem in the language/ecosystem? Lack of knowledge + familiarity
70. How can you tell if it's a problem with the company's internal system or a common problem in the language/ecosystem? We need more content on: DevOps best practices + BEAM
71. How can you tell if it's a problem with the company's internal system or a common problem in the language/ecosystem? Patterns + Anti-patterns · Lessons learned · Benchmarks · Profiling
73. Mapping the challenges Documenting patterns and best practices Sharing real experiences Contributing to the ecosystem
74. And most importantly, we need to train more professionals at EVERY LEVEL
76. A production system involves much more than its code
77. CHALLENGES Code + People + Ecosystem
78. + THE COMMUNITY
79. THE ELIXIR COMMUNITY https://adolfont.medium.com/elixir-brazil-2019-4de3fc06b18f
80. Elixir Meetup São Paulo https://luma.com/elixir_sp
81. https://www.elixirconf.com.br/
82. And it doesn't have to be just about Elixir :)
83. Rails Girls São Paulo https://railsgirls.com.br/
84. Thank you very much speakerdeck.com/elainenaomi linkedin.com/in/elainenaomi illustrations from undraw.co slides available on