Uploaded March 2018 | Updated November 2026, 3 days ago
Register to Lambda Days newsletter to receive updates: lambdadays.org/lambdadays2018/#contact
Rafał Studnicki
Building low tail-latency systems in Erlang and Elixir at Erlang Solutions and Grindr - one of the largest dating apps
Guaranteed SLAs with higher-order functions, and no Magic Numbers!
What if we could apply a combinator to an entire system to ensure that it meets
SLAs? Such a construct exists, and it's called a feeback control loop. By iteratively applying a controller function to the output of our system, we can steer its behavior without any knowledge of the inner workings. This high-level take on Control Theory turns out to be easily implemented in functional languages, such as Elixir.
Because we rely only on functions to build our control schemes, we can easily nest them to manage more that one parameter of the system. This philosophy resulted in a framework that's now used in ensuring the performance of Grindr's production systems in the presence of uncertainty.
For example: if our web service fails to meet the SLA for any reason, the controller function will direct the system to drop a fraction of incoming requests, until all the requests that are accepted, remain within SLA.
lambdadays.org/lambdadays2018/rafal-studnicki
Register to Lambda Days newsletter to receive updates: lambdadays.org/lambdadays2018/#contact
Rafał Studnicki
Building low tail-latency systems in Erlang and Elixir at Erlang Solutions and Grindr - one of the largest dating apps
Guaranteed SLAs with higher-order functions, and no Magic Numbers!
What if we could apply a combinator to an entire system to ensure that it meets
SLAs? Such a construct exists, and it's called a feeback control loop. By iteratively applying a controller function to the output of our system, we can steer its behavior without any knowledge of the inner workings. This high-level take on Control Theory turns out to be easily implemented in functional languages, such as Elixir.
Because we rely only on functions to build our control schemes, we can easily nest them to manage more that one parameter of the system. This philosophy resulted in a framework that's now used in ensuring the performance of Grindr's production systems in the presence of uncertainty.
For example: if our web service fails to meet the SLA for any reason, the controller function will direct the system to drop a fraction of incoming requests, until all the requests that are accepted, remain within SLA.
lambdadays.org/lambdadays2018/rafal-studnicki


![The forgotten advantages of concurrency (Lets #TalkConcurrency - QU2)
Question 2 of Lets #TalkConcurrency - Is there anything forgotten which should be known, or anything which you feel has been overshadowed, which is important [in concurrency]?
Sir Tony Hoare, Joe Armstrong, Carl Hewitt and Francesco Cesarini discuss the positives of concurrency that can otherwise be forgotten...
Full transcript: The forgotten advantages of concurrency (Lets #TalkConcurrency - QU2)](https://i.ytimg.com/vi/uORIxJhOjkI/mqdefault.jpg)







