Inquiry About Impacts of Setting TTL on Live Data in Cosmos DB

Question

We are considering implementing Time-to-Live (TTL) settings on our live data in Azure Cosmos DB to manage data retention more effectively. Before proceeding, we would like to understand the potential impacts this change might have on our query performance and any other aspects of our database operations.

Specifically, we are interested in:

1. How TTL settings might affect query performance, particularly for queries that involve a mix of data with different TTL values.
2. Any potential overhead or performance degradation associated with the automatic deletion of expired items.
3. The impact on indexing and whether TTL settings might influence index efficiency or behavior.
4. Best practices for setting TTL on live data to minimize any negative impacts while achieving our data retention goals.

We would greatly appreciate any insights or recommendations you could provide based on your experience and expertise.

Answer 1

@Malik, Deepankar In addition to the above.

As far as query is concerned, TTL shouldn’t have any significant performance impact. There is a delay between documents getting marked for deletion and the actual deletion. During this delay, the index will still have the terms for the document that have expired. So aggregate queries like COUNT(1) or SUM() which relies on index for the complete evaluation might include the documents that are deleted. This result discrepancy is expected for only a very short period.

Regards,

Oury