
How it works
Indexes are built and are persisted in cheap, cloud storage. The indexes avoid completely duplicating your data by pointing back at your table's base data stored in parquet format - the only data stored by Mach5 is what is required for indexing.
Mach5 indexes data from your databricks delta tables by using your databricks compute clusters.
Mach5's stateless search layer is built as a kubernetes application that serves queries over the data residing in cloud storage.