Optimal Azure Database Development by Karel Coenye
45
Optimal Azure Database Development Karel Coenye et and win an Ignite 2016 ticket #itproceed
-
Upload
itproceed -
Category
Technology
-
view
46 -
download
2
Transcript of Optimal Azure Database Development by Karel Coenye
- 1. Optimal Azure Database Development Karel Coenye Tweet and win an Ignite 2016 ticket #itproceed
- 2. About Me
- 3. As a Database Architect
- 4. The Cloud
- 5. New Application, New possibility Lets go to the cloud!
- 6. First thing First, lets talk about
- 7. Things that cost money Database Troughput Unit (DTU) TPS -> 1 DTU 3320 TPH (Basic) 51,4 TPM (Standard) 0,92 TPS (Premium) Size Time Download Response Time Basic : 80 percentile in 2 sec Standard: 90 percentile in 1 sec Premium: 95 percentile in ,5 sec
- 8. DTU Average - Rules 1 User = 5 DTU (if he would use all the power) User pacing estimated at 1sec 1 User => 1 DTU realistically Tiers are size, Response Time and DTU based Performance has to fit in the tier, otherwise it becomes to expensive
- 9. DTU Explained Read Lite SELECT; in-memory; read-only 35 % Read Medium SELECT; mostly in-memory; read-only 20 % Read Heavy SELECT; mostly not in-memory; read-only 5 % Update Lite UPDATE; in-memory; read-write 20 % Update Heavy UPDATE; mostly not in-memory; read-write 3 % Insert Lite INSERT; in-memory; read-write 3 % Insert Heavy INSERT; mostly not in-memory; read-write 2 % Delete DELETE; mix of in-memory and not in-memory; read-write 2 % CPU Heavy SELECT; in-memory; relatively heavy CPU load; read-only 10 % The overall mix has a read/write ratio of approximately 2:1
- 10. How design can save cost Limit heavy reads Limit heavy writes Limit CPU intensive queries Limit Useless Cycles Limit the result sets Limit Contention lets focus on architecture
- 11. Stevie Wonder goes to a concert
- 12. Common Cloud Issues Availability Data Management Design And Implementation Messaging Management and Monitoring Performance And Scalability Resiliency Security
- 13. Management and Monitoring
- 14. External Configuration Store
- 15. Metering
- 16. Stevie Wonder wants to sing
- 17. Performance And scalability
- 18. Stateful vs. Stateless
- 19. Think outside the box
- 20. Segregation Of Data
- 21. Small High-Performance OLTP Small Stateful Strictly Consistent High Performance Scales via Elastic Databases Contains minimal operational dataset
- 22. Large Single Version of the truth Datastore Contains the truth Contains all the data Highly Normalized (3Rd NF or Higher) Stateful Transactionally Consistent Mix of Strict and eventually consistent Master of the Data
- 23. Read-Only / Reporting Warehouse / Sync DB (semi)-Stateless Contains subset of data Has eventual consistency Scales via sync/replication
- 24. Competing Consumers
- 25. Cache-Aside
- 26. Remote Computing Data Data Cache Stored Procedures Data Cache Stored Procedures Data Cache Stored Procedures
- 27. Design and Implementation
- 28. Command and Query Responsibility Segregation
- 29. Materialized Views
- 30. Managing Data Consistency
- 31. Strong Consistency
- 32. Eventual Consistency
- 33. Replicating And Synchronizing
- 34. Replicating Data(bases)
- 35. Master - Master
- 36. Master - Slave
- 37. Replication
- 38. Synchronizing Data(bases)
- 39. Sync Works on
- 40. Data Sync
- 41. How does Stevie sound at the Metallica Concert?
- 42. And win a Lumia 635 Feedback form will be sent to you by email Give me feedback
- 43. Follow Technet Belgium @technetbelux Subscribe to the TechNet newsletter aka.ms/benews Be the first to know
- 44. Thank you!
- 45. Belgiums biggest IT PRO Conference