Pony: How I learned to stop worrying and embrace an unproven technology

This is the "Fintech" version of this talk.

Abstract:

Conventional wisdom holds that you should never build your business on top of bleeding-edge technology. Yet, when we at Sendence set out to build our high-performance stream processing engine Wallaroo, we chose to build it in Pony- a high-performance actor-based programming language that has limited industry usage. So far, this decision has worked out very well for us. Pony has helped us move with speed and confidence. In this talk, I'll cover several Pony features that have contributed to moving us forward. I'll also discuss how Pony's immaturity has at times hindered us. By the end of this talk, you'll have a solid understanding of Pony's strengths and weakness and be better equipped to decide if it just might be the right language for your next project.

Transcript

1. Pony
   How I learned to stop worrying and embrace an unproven technology
   

   View Slide

2. Sean T Allen
   Author of Storm Applied
   Member of the Pony core team
   VP of Engineering at Sendence
   

   View Slide

3. Pony
   Open-source, object-oriented,
   actor-model, capabilities-
   secure, high-performance
   programming language
   

   View Slide

4. This talk is…
   The story of Pony told through the story of Wallaroo;
   the product we built with it…
   

   View Slide

5. Wallaroo
   High-performance, distributed
   stream processor
   High-throughput
   Low-Latency
   Consistent performance
   Resource efficient
   

   View Slide

6. -Sylvan Clebsch
   “
   A programming language is just
   another tool. It’s not about syntax.
   It’s not about expressiveness. It’s
   not about paradigms or models.
   It’s about managing hard
   problems.”
   

   View Slide

7. Why Pony?
   

   View Slide

8. Why Pony?
   Highly concurrent
   Predictable latency demands
   Data Safety
   “Batteries” not required
   

   View Slide

9. Highly Concurrent
   

   View Slide

10. Highly Concurrent
    Actor model
    Async messaging
    Work stealing scheduler
    Mechanical sympathy
    

    View Slide

11. Predictable latencies
    

    View Slide

12. Predictable Latencies
    No “stop the world” GC
    Per-actor heaps
    Better clustering
    

    View Slide

13. Data Safety
    

    View Slide

14. Data Safety
    Reference Capabilities
    

    View Slide

15. Batteries not required
    

    View Slide

16. Batteries not required
    Small Standard Library
    C-FFI
    

    View Slide

17. Results
    

    View Slide

18. Results
    18 month jumpstart for “free”
    Excellent performance
    More confident refactoring
    Some pain
    

    View Slide

19. Pain
    

    View Slide

20. Pain
    Pre-1.0
    Compiler bugs
    Breaking changes
    Limited tooling
    Runtime knowledge required
    

    View Slide

21. Is Pony right for you?
    

    View Slide

22. Is Pony right for you?
    You have a hard concurrency
    problem
    You aren’t reliant on a lot of
    existing libraries
    You are willing to write “most
    everything” from scratch
    Yes, if…
    

    View Slide

23. “Fintech” Questions
    

    View Slide

24. “Fintech” Questions
    How do I shard my data?
    How do I do priority queues?
    How do I only process the most
    recent messages?
    

    View Slide

25. Takeaways…
    Pony has a powerful, data-race free,
    concurrency-aware type system
    
    The Pony runtime can help you solve
    hard concurrency problems
    
    You might be able to use Pony for
    fintech in production now
    

    View Slide

26. Trash Day paper:
    https://www.usenix.org/system/files/conference/hotos15/
    hotos15-paper-maas.pdf
    ORCA paper:
    https://www.ponylang.org/media/papers/OGC.pdf
    

    View Slide

27. @seantallen
    www.monkeysnatchbanana.com
    @SendenceEng
    www.sendence.com
    @ponylang
    www.ponylang.org
    #ponylang on freenode
    

    View Slide