Data Modeling Decisions for Azure Cosmos DB (Cosmos DB Conf 2026)
Hasan Savran explains how early data modeling choices in Azure Cosmos DB affect scalability, query performance, and cost, and what to decide up front to avoid painful changes later.
Full summary based on transcript
What the session focuses on
The session is about making practical data modeling decisions for Azure Cosmos DB, with emphasis on choices that are hard to change later and that directly impact:
- Application scalability
- Query performance
- Cost (through throughput and query patterns)
Choosing a partition key
Hasan Savran covers strategies for selecting a partition key that supports expected access patterns and growth, including:
- Picking a key that distributes data and workload effectively
- Avoiding designs that create hot partitions
- Considering how partitioning impacts query patterns and throughput usage
Schema design tradeoffs
The talk discusses tradeoffs in schema design for Cosmos DB, including how different modeling approaches can affect:
- Read vs write efficiency
- Query complexity
- Future flexibility when requirements change
Estimating storage and scaling needs
The session highlights planning for growth by estimating:
- Storage requirements
- Scaling needs as data volume and traffic increase
- How early modeling decisions can lead to unexpected resource requirements later
Managing cross-partition queries
Hasan Savran explains how to think about cross-partition queries and how to manage them efficiently, including:
- Understanding when queries will fan out across partitions
- Designing data and queries to reduce cross-partition overhead
- Considering the cost/performance impact of cross-partition access patterns
Links and references
- Cosmos Conf 2026 playlist: https://aka.ms/CosmosConf26Playlist
- Cosmos Conf Challenge (DP-420 voucher details): https://aka.ms/CosmosDBConfChallenge
- Post-event survey: https://aka.ms/CosmosConf2026Survey
- Conference site: https://aka.ms/azurecosmosdbconf
- Speaker links: