In his article, Zachary Tong explores various methods for managing relational data within Elasticsearch, emphasizing its inherently flat structure. He discusses three primary approaches: inner objects, nested types, and parent/child relationships, each with distinct benefits and limitations. Inner objects are simple and efficient but only suitable for one-to-one relationships. Nested types offer improved query accuracy by maintaining independent nested documents, though they come with performance costs due to forced reindexing during updates. Parent/child relationships allow separate storage of related documents, facilitating easier updates but at the expense of slightly reduced performance and complex sorting or scoring. Tong highlights denormalization as a flexible alternative that involves manually managing relationships, offering potential power and flexibility, albeit with increased administrative effort. Through these insights, Tong provides a comprehensive overview of how each method can be applied based on specific data requirements and usage scenarios in Elasticsearch.