LFC Photon Engine Optimization 4

Question 1

A data engineer notices that a Photon-enabled query has low CPU utilization but high I/O wait. What is the most likely bottleneck?

Accepted Answer

The query is I/O bound due to reading many unoptimized small files or non-columnar data

Answer

High I/O wait with low CPU suggests the data access pattern—not Photon's compute—is the bottleneck, often caused by small files or row-based formats.

Question 2

What is the effect of enabling liquid clustering on a Delta table used frequently with Photon?

Accepted Answer

Liquid clustering continuously reorganizes data to improve file-level statistics and Photon's data skipping

Answer

Liquid clustering automatically maintains data co-location over time, improving the selectivity of Delta's file statistics that Photon uses for data skipping.

Question 3

In which Databricks Runtime version was Photon first made generally available?

Accepted Answer

DBR 9.1 LTS

Answer

Photon became generally available in Databricks Runtime 9.1 LTS, marking its production-ready release.

Question 4

Which join strategy does Photon optimize most aggressively for large-to-large table joins?

Accepted Answer

Sort-merge join with native vectorized sorting and probing

Answer

Photon's native sort-merge join implementation uses vectorized sorting and hash probing for large equi-joins where broadcasting is not feasible.

Question 5

What is the recommended approach when a specific operator in a query falls back from Photon to JVM Spark?

Accepted Answer

Identify the unsupported operator and rewrite the query to use a Photon-supported equivalent

Answer

Identifying and rewriting the unsupported operator (e.g., replacing a Python UDF with a built-in SQL function) allows the full query to benefit from Photon.

(LFC) Lakehouse Fundamentals Certification Practice Test

(LFC) Lakehouse Fundamentals Certification Practice Test

LFC Photon Engine Optimization 4