chronix.spark
chronix.spark copied to clipboard
ChronixDB
Time series analysis with Apache Spark based on Chronix |
- Sub-chunking of a time series in fragments of a chronix chunk size - Linear regression on chunks - Mapping of regression values to symbolic representations
- join (chunks) - diff / delta - filter - lag - align - vectorization - statistics based on DoubleRDD incl. approx. - be inspired by - Spark-TS: TimeSeriesRDD -...
- Allow serialization in memory or on disk - Use Kryo as tried before or use another format like ByteArray or Apache Arrow. `conf.set("spark.serializer", "org.apache.spark.serializer.KryoSerializer"); conf.registerKryoClasses(new Class[]{ MetricTimeSeries.class, MetricObservation.class, MetricDimensions.class,...
Provide bidi transformations from/to ChronixRDD and Spark-TS' TimeseriesRDD
- separate project - candidates: DoubleList, Vector (generify MetricVector and maybe introduce char[][] instead of String[] and use dictionary) - use them in BinaryTimeSeries