Dremio hosted the first Subsurface conference on cloud data lake use cases and architectures in mid-August. In this post, Paul Singman summarizes a few talks from the conference. Tomer Shiran, CPO of Dremio, discussed how data architectures have evolved from monoliths centered on a proprietary database or Hadoop clusters to more decoupled systems that include separate storage, data, compute, and client layers. Wes McKinney, Director of Ursa Labs, described how Apache Arrow can facilitate the integration of disparate data handling systems, obviating the need for complex custom data connectors. Maxime Beachemin, CEO of Preset, explored how functional programming principles can be applied to build more reliable data pipelines.