Centralized Counters in a Distributed World - Event-Based Realtime Statistics using eTally

When running large systems monitoring throughput is important. Most solutions do this by counting events in buckets of days or hours. this approach has the advantage of being fast, and taking up a very limited amount of space.

eTally's approach is different in that it tracks moving windows and keeps dynamically updated leaderboards for events at different time resolutions, making all search queries to the system run in O(1) time and space, and making event handling in O(#containing_leaderboards * ln(size_of_leaderboard) ) time and O(1) space.

Talk objectives:

The goal would be to introduce eTally, and explain the structural techniques that make the solution efficient.
Show how to leverage other languages from Erlang and make awesome sollutions in the process. 

Target audience:

Engineers designing high throughput system, users of the technology, and algorithmic/data structure interested people. 

Video

Slides
Video

About Martin

Martin is a senior Erlang engineer at Tigertext, with a masters degree is computer science from the university of Aarhus Denmark. Previously he worked in machine learning and medical imaging.

Twitter: @mskristiansen

Github: martinsk

Back to conference page