My summary/advices based on the video (can be a TL;DR for you): 1. Choose partition keys thoughtfully, suitable for your data volume and write/access patterns. Some issues don't appear with low amount of data, don't make conclusions too early. Learn about physical partitions. 2. Understand consistency models and choose thoughtfully. It all comes down to "when your writes become available to your reads". The higher consistency you need, the lower latency and throughput you get. Most use-cases don't require Strong Consistency. Use Eventual until you have a reason not to. 3. Queries: Query only on indexed data. Prefer in-partition queries over cross-partition queries. Check index metrics (SDK) and query stats (web UI). 4. Cosmos is a row store. Do column-oriented operations (avg, sum, etc.) on an analytical store, like Azure Synapse Analytics, i.e. try Azure Synapse Link. 5. SDK issues. Share one connection for the entire application (have a single CosmosClient instance). Read the documentation on how to use it correctly. But as always: read documentation, the more the better.
Can we get shorter summary videos of these? Or at least a table of contents with timestamps? 1 and a half hours is a bit too long to watch, but I want to learn about antipatterns.
My summary/advices based on the video (can be a TL;DR for you):
1. Choose partition keys thoughtfully, suitable for your data volume and write/access patterns. Some issues don't appear with low amount of data, don't make conclusions too early. Learn about physical partitions.
2. Understand consistency models and choose thoughtfully. It all comes down to "when your writes become available to your reads". The higher consistency you need, the lower latency and throughput you get. Most use-cases don't require Strong Consistency. Use Eventual until you have a reason not to.
3. Queries: Query only on indexed data. Prefer in-partition queries over cross-partition queries. Check index metrics (SDK) and query stats (web UI).
4. Cosmos is a row store. Do column-oriented operations (avg, sum, etc.) on an analytical store, like Azure Synapse Analytics, i.e. try Azure Synapse Link.
5. SDK issues. Share one connection for the entire application (have a single CosmosClient instance). Read the documentation on how to use it correctly.
But as always: read documentation, the more the better.
Great content, learned a lot. You guys can think about TLDR version :-)
Can we get shorter summary videos of these? Or at least a table of contents with timestamps? 1 and a half hours is a bit too long to watch, but I want to learn about antipatterns.