GraphFrames provides a unified API for graph queries and algorithms in Spark SQL. It translates graph patterns and algorithms to relational operations optimized by the Spark SQL query optimizer. Materializing the right views, such as the triplet view for GraphX algorithms or user-defined views for queries, can improve performance. An evaluation shows GraphFrames outperforms Neo4j for unanchored queries and approaches GraphX performance for graph algorithms using Spark SQL. Future work includes automatically suggesting optimal views and exploiting attribute-based partitioning.