I'm diving into DynamoDB and NoSQL architecture, and I'm looking for advice on how to best structure my keys for efficiency. I want to minimize the use of scans and focus on querying instead. I have a set of records with many-to-many relationships. The documentation suggests a design like this:
pk sk attributes
--------------------------------------
Parent#123 Parent#123 {parent details}
Parent#123 Child#456 {child details}
I'm building an API to list all parents. How can I query this setup without using a scan? Currently, my pk/sk design looks like this:
pk sk attributes
--------------------------------------
Parent Parent#123 {parent details}
Parent#123 Child#456 {child details}
This allows me to query using the pk 'Parent', but I need to ensure key integrity when adding Child records. Any advice?
6 Answers
Using a GSI could be key here! Consider modeling your data to include a type field that distinguishes between parents and children for easier querying.
You might consider implementing a Global Secondary Index (GSI). This index could help you create a parent lookup similar to your current design, but note that this will require double writes to maintain consistency.
Just a thought: make sure a NoSQL solution fits your use case. Start with your data access requirements, then consider whether NoSQL is the right type of database for your project.
Take some time to outline all the necessary queries upfront. This will help in determining the proper table keys and any indexes you may need. Prioritize the most commonly used queries for optimal performance.
Do you have an estimate of how many parent items will be returned? If it's only a few hundred, that might simplify the structure.
Check out Alex DeBrie’s book on DynamoDB patterns; it’s really helpful! For managing key constraints for related records, you can handle that in your application or use condition expressions in a transaction. Here are some resources:
1. [Alex DeBrie's Guide](https://www.alexdebrie.com/posts/dynamodb-condition-expressions/#1-confirming-existence-or-non-existence-of-an-item)
2. [Transactional Write Items](https://docs.aws.amazon.com/amazondynamodb/latest/APIReference/API_TransactWriteItems.html)
Absolutely, leveraging transactional writes sounds like the right path for you!
It's crucial to model your pk/sk based on how you'll be accessing your data. Consider the types of queries you’ll need, like listing all parents or finding children for a specific parent. This way, you can avoid scans effectively.