Developers are increasingly adopting continuous profiling to monitor and debug application performance issues. However, continuous profiling produces high-volume data that is difficult to store and query efficiently. To enable the analysis of profiling data and support other write-heavy applications that may read lots of data at once, Polar Signals has OSS’ed FrostDB. FrostDB is an embeddable columnar database, implemented in Go and based on Apache Parquet and Arrow. Unlike other embeddable columnar databases, PolarDB supports dynamic columns so it can translate between a label-based data model and a table-based layout efficiently.