Techdays NL - Azure Cosmos DB - a technical overview

https://www.youtube.com/watch?v=-23wqlIf3nM

Transcript

1. Azure Cosmos DB A technical overview @pascalnaber
2. Multiple consistency models Guaranteed: Low latency High availability Predictable throughput Schema agnostic Elastic horizontally scale out Multi model Globally distributed
3. NoSQL SQL Azure MySQL Azure PostgreSQL Azure Table Storage Azure Cosmos DB
4. Unstructured data No joins Schema free Huge amount of data Linear scalable Eventual Consistent
5. Project Florence DocumentDB Cosmos DB 2010 2015 10 May 2017 Founder: Dharma Shukla
6. Globally distributed
7. Globally distributed
9. Azure Cosmos DB Key-Value Column-Family Graph Documents SQL DocumentDB Azure Tables
10. SQL (DocumentDB) & MongoDB Table (key-value) Gremlin (graph)
11. Database account Database User Permission {Container} {Item} sproc trigger UDF conflict attachment Collection Graph Table Document Node Row Edge
12. • Key • PartitionKey • RowKey • Value • Properties TableEntity WindowsAzure.Storage-PremiumTable WindowsAzure.Storage
13. Microsoft.Azure.DocumentDB MongoDB.Driver
14. Edge Node / Vertex Microsoft.Azure.Graphs Microsoft.Azure.DocumentDB Properties Properties
15. Pascal Dumb & Dumber Toy Story Kelsie Jumanji Vertex Vertex Edge Edge Jeff Daniels
16. g.V('user7').outE('rates').has('weight', gte(4.5)).inV().inE('rates').has('weight', gte(4.5)).outV().outE('rates').has('weight', gte(4.5)).inV()
17. The most important design decision for your Cosmos DB project
20. PartitionKey Read Write Postcode* Guid BlogId* ArticleGroupId*
23. Reads are guaranteed to see the most recent write Single location only! Consistency levels 1 2 3 4 5 6 7 8 9 10 Total Britt 0 1 4 2 7 Maud 0 2 3 5 Strong Consistency 7-5 Bounded Staleness (lag 5 min) 5-5, 7-5 Session (self) 7-5 Session (anybody else) 0-0, 0-2, 1-0, 1-2, 5-2, 1-5, 5-5, 7-5 Consistent Prefix 0-0, 1-0, 1-2, 5-2, 5-5, 7,5 Eventual Consistency 0-0, 0-2, 1-0, 1-2, 5-2, 1-5, 5-5, 7-5
24. Reads can lag behind writes by a user specified number of versions or time interval High Latency Consistency levels 1 2 3 4 5 6 7 8 9 10 Total Britt 0 1 4 2 7 Maud 0 2 3 5 Strong Consistency 7-5 Bounded Staleness (lag 5 min) 5-5, 7-5 Session (self) 7-5 Session (anybody else) 0-0, 0-2, 1-0, 1-2, 5-2, 1-5, 5-5, 7-5 Consistent Prefix 0-0, 1-0, 1-2, 5-2, 5-5, 7,5 Eventual Consistency 0-0, 0-2, 1-0, 1-2, 5-2, 1-5, 5-5, 7-5
25. Read My Writes Low Latency Consistency levels 1 2 3 4 5 6 7 8 9 10 Total Britt 0 1 4 2 7 Maud 0 2 3 5 Strong Consistency 7-5 Bounded Staleness (lag 5 min) 5-5, 7-5 Session (self) 7-5 Session (anybody else) 0-0, 0-2, 1-0, 1-2, 5-2, 1-5, 5-5, 7-5 Consistent Prefix 0-0, 1-0, 1-2, 5-2, 5-5, 7,5 Eventual Consistency 0-0, 0-2, 1-0, 1-2, 5-2, 1-5, 5-5, 7-5
26. Readers will never see writes that are out of order Low latency Consistency levels 1 2 3 4 5 6 7 8 9 10 Total Britt 0 1 4 2 7 Maud 0 2 3 5 Strong Consistency 7-5 Bounded Staleness (lag 5 min) 5-5, 7-5 Session (self) 7-5 Session (anybody else) 0-0, 0-2, 1-0, 1-2, 5-2, 1-5, 5-5, 7-5 Consistent Prefix 0-0, 1-0, 1-2, 5-2, 5-5, 7,5 Eventual Consistency 0-0, 0-2, 1-0, 1-2, 5-2, 1-5, 5-5, 7-5
27. Out of order reads At the end it will be fine lowest read and write latency Consistency levels 1 2 3 4 5 6 7 8 9 10 Total Britt 0 1 4 2 7 Maud 0 2 3 5 Strong Consistency 7-5 Bounded Staleness (lag 5 min) 5-5, 7-5 Session (self) 7-5 Session (anybody else) 0-0, 0-2, 1-0, 1-2, 5-2, 1-5, 5-5, 7-5 Consistent Prefix 0-0, 1-0, 1-2, 5-2, 5-5, 7,5 Eventual Consistency 0-0, 0-2, 1-0, 1-2, 5-2, 1-5, 5-5, 7-5
28. & reserve
29. Cosmos DB Pricing (per month) RU’s p/s € 400 20,08 1.000 50,19 5.000 250,97 10.000 501,94 50.000 2509,70 100.000 5019,40 GB € 1 0,21 10 2,11 50 10,55 100 21,09 1.000 210,90 10.000 2109,00 1 KB read 1 KB write Select by id Select by xyz Select group by Select top 10 = 1 RU = 5 RU = 2.5 RU = 7 RU = 70 RU = 10 RU
30. https://aka.ms/cosmosdb-emulator https://azure.microsoft.com/en-us/try/cosmosdb/ https://www.documentdb.com/capacityplanner docker pull microsoft/azure-cosmosdb-emulator
33. Json MongoDB SQL Server Csv files Azure Table Storage Amazon DynamoDB Hbase Every four hours Manual backups Backups
34. Multiple consistency models Guaranteed: Low latency High availability Predictable throughput Schema agnostic Elastic horizontally scale out Multi model Globally distributed
35. Thank you! Pascal Naber Coding cloud architect Xpirit @pascalnaber
36. pnaber@xpirit.com @pascalnaber http://www.xpirit.com https://pascalnaber.wordpress.com