What an excellent video. The first Microsoft employee that I've come across that didn't define cosmos DB in marketing terms, and instead said what it actually is; a no SQL database on Azure. The rest of the presentation was fantastic as well.
Best explanation of applying partition keys and how it impacts RUs I have watched - and I have watched quite a few here and on PluralSight. I finally feel like I understand these areas fully and can move ahead with refactoring my SQL project.
If you're refactoring a relational database then you definitely should watch this video where Mark Brown refactors Adventure Works from SQL DB to Cosmos DB. czcams.com/video/3gEytt-Jops/video.html
You don't need to do that everytime. Everything has trade-offs. What the speaker is trying to convey is that you should measure whether it makes sense to denormalize or not.
I think the main point is that typical SQL joins/constrains are impossible(or impractical) IF you care about horizontal scalability and good response times. The speaker makes a good case for denormalization, but that is not a silver bullet obviously. But it is scalable if you can have unbounded storage and keep consistent low response times.
exactly. Imagine every time a customer places an order we run that javascript SP to update the order count. How computationally intensive is that. Writes are gonna be slow.
@@AzureCosmosDB He seemed to have started with saying whether you embed or not depends on the nature of the relationship, which is right but then proceeds to change his model and embed everything because of what seems like a limitation of the tool (Cosmos with SQL API I assume). At MongoDB, they actually follow up by showing how to join and retrieve from different collections. I imagine the MongoDB for CosmosDB now has that feature?
I thought that with persistent memory technology there would be a cosmos database SKU to put the entire database in memory so one doesn’t need to do all this complicated modeling.
One of the best videos about Cosmos DB! Not just sales pitch, but the real use cases.
What an excellent video. The first Microsoft employee that I've come across that didn't define cosmos DB in marketing terms, and instead said what it actually is; a no SQL database on Azure.
The rest of the presentation was fantastic as well.
Glad you enjoyed it!!!
Best explanation of applying partition keys and how it impacts RUs I have watched - and I have watched quite a few here and on PluralSight. I finally feel like I understand these areas fully and can move ahead with refactoring my SQL project.
If you're refactoring a relational database then you definitely should watch this video where Mark Brown refactors Adventure Works from SQL DB to Cosmos DB. czcams.com/video/3gEytt-Jops/video.html
Probably the best video out in internet on real time example
Loved your explanations on everything! Cosmos DB will be my first DB to work with, and I am even more excited after watching this video!
In reality , how do u load bulk data . Will u be making all your data in json and load ?
great explaination video, thank you!
What an explanation. Hats off!
Crystal clear explanation. It helped me on Document DB Design.
Lovely explanation on Cosmos DB Design. Thanks bro!
Glad it was helpful!
Excellent video...really helped me in a design work
This is great!
Can a logical partition span across more than one physical server ?
No it can't. By design logical partitions reside within one physical partition.
A logical partition can max 20GB not 10 GB
This video is from 2019, before the 20 GB limit was introduced.
So every time there is a query requirement, you denormalise and store that value in the document? How is that scalable :/
You don't need to do that everytime. Everything has trade-offs. What the speaker is trying to convey is that you should measure whether it makes sense to denormalize or not.
I think the main point is that typical SQL joins/constrains are impossible(or impractical) IF you care about horizontal scalability and good response times. The speaker makes a good case for denormalization, but that is not a silver bullet obviously. But it is scalable if you can have unbounded storage and keep consistent low response times.
exactly. Imagine every time a customer places an order we run that javascript SP to update the order count. How computationally intensive is that. Writes are gonna be slow.
@@AzureCosmosDB He seemed to have started with saying whether you embed or not depends on the nature of the relationship, which is right but then proceeds to change his model and embed everything because of what seems like a limitation of the tool (Cosmos with SQL API I assume). At MongoDB, they actually follow up by showing how to join and retrieve from different collections. I imagine the MongoDB for CosmosDB now has that feature?
I thought that with persistent memory technology there would be a cosmos database SKU to put the entire database in memory so one doesn’t need to do all this complicated modeling.