Everything is a time-series. Where are the corresponding abstractions? How to …
summary
Concrete claim: time series are modelled as functions-of-time where each sample carries a validity duration, not just a timestamp, and values are step functions so no interpolation changes the past. The library scala-timeseries-lib exposes a merge operator that takes (Option, Option) => Option to combine misaligned, sparse, heterogeneous series; it powers wind‑park asset monitoring, serialises triples to S3 or RocksDB, publishes to Maven Central, fits in RAM today, and is reported to simplify real deployments.
AI-written summary · about
chapters (13)
- 0:00Intro
- 1:01Contents
- 2:34What is a timeseries
- 4:30Complex timeseries
- 6:33Why would we do this
- 7:39Why a new library
- 8:36Assumptions
- 9:26No interpolation
- 10:56Realtime vs analytics
- 12:41In an ideal world
- 13:24Realtime examples
- 14:30Plumbing
- 17:35Questions
related talks
01h 43m(Serious) Time for Time Series
01h 27mIntroduction to Machine Learning with Time Series
45mUsing time-series data to better understand our products & infrastructure
26mImproving Observability in Your Observability: Simple Tips for SREs
38mAdd ALL the things: abstract algebra meets analytics
32mWhy Use Scala in a Bigdata Context?
24mLightning Talks
19mContinuous Arrays