Author: Wenbo Zhao
Presented at: Spark Summit 2016, San Francisco, CA
Abstract: The ability to analyze time series of data at scale is critical for the success of finance and IoT applications based on Spark. This session introduced Huohua, Two Sigma’s implementation of highly optimized time series operations in Spark. Huohua performs truly parallel and rich analyses on time series data by taking advantage of the natural ordering in time series data to provide locality-based optimizations. Huohua is an open source library for Spark based around the OrderedRDD, a time series aware data structure, and a collection of time series utility and analysis functions that use OrderedRDDs. Unlike DataFrames and Datasets, Huohua’s OrderedRDDs can leverage the existing ordering properties of datasets at rest and the fact that almost all data manipulations and analysis over these datasets respect their ordering properties. It differs from other time series efforts in Spark in its ability to efficiently compute across panel data or on large scale high frequency data. In this talk, we will present the architecture of OrderedRDDs and its integration with Spark SQL, DataFrames, and Datasets; and the analysis tools we’ve built on top to merge, join, aggregate, and intervalize data, and compute windowed, rolling, and cycle-based summarizations and cross-panel analysis. We’ll present results comparing Huohua to similar operations using off the shelf RDDs and DataFrames.