Photon

Photon delivers an additional 2x speedup over the TPC-DS 1TB benchmark compared to the latest Databricks Runtime (DBR) versions, and customers have observed 3x–8x speedups on average depending on their workloads. Photon is also cited as enabling up to 12x speedups and up to 12x better price/performance compared to other cloud data warehouses. These gains come from Photon's vectorized, C++-based execution engine, which takes advantage of modern hardware.

No — Photon is an ANSI-compliant engine designed to be compatible with modern Apache Spark DataFrame and SQL APIs, so existing SQL, Python, R, Scala and Java code runs without changes. All you have to do to benefit from Photon is turn it on; it seamlessly coordinates work and resources and transparently accelerates portions of your queries with no tuning or manual intervention required. This design also avoids vendor lock-in since workloads remain standard Spark.

Photon is built for data teams who want to standardize on one engine and one set of APIs across ETL, analytics and data science workloads, whether running in batch or streaming. Because it requires no code changes, it suits teams already running Apache Spark or Databricks Runtime who want faster performance without migrating to new tooling. It is also relevant for teams running large-scale production jobs on SQL and Spark DataFrames who need lower latency and cost at scale.

Photon accelerates a range of workloads including large-scale production jobs on SQL and Spark DataFrames, faster time-series analysis for IoT applications, and querying petabyte-scale data sets for data privacy and compliance needs such as identifying and deleting records without duplicating data (used together with Delta Lake). It also speeds up loading data into Delta Lake and Parquet tables through vectorized I/O, lowering the overall runtime and cost of data engineering jobs.

The page does not list a separate price for Photon; instead, it highlights the cost impact of enabling it, citing up to 80% total cost of ownership (TCO) savings and up to 12x better price/performance compared to other cloud data warehouses. These savings stem from Photon's faster query execution, which reduces the compute time needed to run the same workloads. Because Photon works within the Databricks Runtime you already use, evaluating exact cost impact for your workloads is best done by trying it directly on your own jobs.

Getting started with Photon is as easy as turning it on, since it requires no code changes and works with your existing Apache Spark SQL and DataFrame code. From there, Photon automatically coordinates work and resources to accelerate eligible portions of your queries without manual tuning. You can begin exploring Photon by signing up through Try Databricks for free.