You can't access the Schema Advisor for serverless instances.
You can model your schema based on frequently used design patterns. The Building with Patterns blog series discusses the following frequently used design patterns.
To read about situations in which arrays work well, see the following design patterns:
Use The Attribute Pattern for handling data with unique combinations of attributes, such as movie data where each movie is released in a subset of countries.
Use The Bucket Pattern for handling tightly grouped or sequential data, such as time span data.
Use The Polymorphic Pattern for handling differently shaped documents in the same collection, such as athlete records across several sports.
To read about strategies for keeping documents in your working set at a manageable size, see the following patterns:
Use The Extended Reference Pattern to duplicate a frequently-read portion of data from large documents to smaller ones.
Use The Subset Pattern to reduce the size of documents with large array fields.
Use The Outlier Pattern to handle a few large documents in an otherwise standard collection.
To learn how to incorporate the flexible data model into your schema, see the following presentations from MongoDB.live 2020:
Atlas offers two ways to detect common schema design issues and suggests modifications that follow MongoDB’s best practices:
The Performance Advisor provides holistic schema recommendations for your cluster by sampling documents in your most active collections and collections with slow-running queries.
The Atlas UI offers schema suggestions for a specific collection by sampling documents in that collection.
To learn more about how to apply the suggestions offered in either the Performance Advisor or Data Explorer, refer to the following pages:
Reason for Suggestion
You are running too many
Your documents contain array fields with many elements, which can degrade query performance.
You have unnecessary indexes in your collection, which can consume disk space and degrade write performance.
You have excessively large documents, which can degrade the performance of your most frequent queries.
You have an exceedingly high number of collections in a database, which can result in unnecessary disk space usage.
Schema suggestions for a collection are partly driven by a random sampling of documents from that collection. Because this sampling is performed each time your schema is analyzed, you may see different suggestions at different times for the same collection.
The Performance Advisor monitors slow queries recognize certain schema issues, namely too many
$lookupoperations and not utilizing an index for case-sensitive regex queries. If a cluster does not consistently receive long-running queries, the Performance Advisor may not suggest all potential improvements for that cluster, or may not show all reasons why an improvement is being suggested.