7 Key Things for Building a Highly-Scalable SharePoint 2013 App
Download as PPTX, PDF4 likes2,509 views
The document outlines seven key strategies for building a highly scalable SharePoint app, emphasizing the importance of effective caching, content delivery networks, and storage solutions. It discusses the need for redundancy, queuing mechanisms, and asynchronous processing to ensure performance under heavy user loads. The insights aim to help developers navigate the challenges of scalability in SharePoint applications.
7 Key Things for Building a Highly-Scalable SharePoint 2013 App
- 1. 7 Key Things for Building a Highly-Scalable SharePoint App Edin Kapić – SharePoint Architect, Spenta
- 2. Edin Kapić www.sharepointsummit.org SharePoint Architect Works for Spenta / Beezy in Barcelona SharePoint Server MVP President of Catalonian SharePoint User Group (SUG.CAT) @ekapic www.edinkapic.com sug.cat 2
- 4. Scalability www.sharepointsummit.org Hey boss, we have 1M new users for our web app!! ...and they all sign up tomorrow!
- 9. SharePoint 2013 Application Architecture www.sharepointsummit.org SharePoint 2013 Apps are cloud-based
- 10. SharePoint 2013 Application Architecture www.sharepointsummit.org SharePoint is now just another external system Mechanisms to minimize round trips 10
- 11. SharePoint 2013 Application Architecture www.sharepointsummit.org With the cloud, we have potential for scalability and high availability But the architecture must enable it (It doesn’t come out of the box) 11
- 12. SharePoint 2013 Application Architecture www.sharepointsummit.org 12
- 13. Session Overview www.sharepointsummit.org Avoiding Roundtrips Effective Caching Mechanisms Content-Delivery Networks Avoiding Bottlenecks Avoiding Storage Locks Waiting in a Queue is Not Bad Acting Asynchronously Avoiding Single Point of Failure Redundant Design 13
- 14. SharePoint 2013 Application Architecture www.sharepointsummit.org 14
- 16. Effective Caching Mechanisms www.sharepointsummit.org Caching is the cheapest way to avoid unnecessary roundtrips Local Cache Small, static data, very frequently accessed Distributed Cache Everything that you reasonably can Stale Data
- 18. BLOB Storage www.sharepointsummit.org Big binary storage in the cloud Bypass the Processing Public BLOBs can be accessed by URL Shared Signatures Private BLOBs can be accessed by special URL
- 19. Content Delivery Networks (CDN) www.sharepointsummit.org Big binary storage close to the user No need to serve from the cloud First user pays the “toll”, others get it free Offload all that you can Images Scripts Media Files Versioned URLs
- 21. Avoiding Storage Locks www.sharepointsummit.org NoSQL vs RDBMS Table Storage vs SQL Azure Eventual vs Immediate Consistency
- 22. Avoiding Storage Locks www.sharepointsummit.org Partition Keys Sharding Partition your data across multiple databases or storages A-F G-O JOINs across shards Data Access SQL Azure Federations P-Z Shards
- 23. Avoiding Storage Locks www.sharepointsummit.org Command-Query Responsibility Segregation Source: Udi Dahan 23
- 24. www.sharepointsummit.org Waiting in a Queue is Not Bad
- 25. Waiting in a Queue is Not Bad www.sharepointsummit.org Request/Response doesn’t scale well Web Back-End Message Queue By queuing the requests we have control To decouple the request and the response To do retries and accomodate node loss To throttle the rate of service To scale up and down depending on the workload
- 26. Waiting in a Queue is Not Bad www.sharepointsummit.org Azure Queues Low-Level Pass messages between roles and sites Fully decoupled R/W operations Automatic Load Balancing Azure Service Bus Queues High-Level Messaging framework with queues Publisher-Subscriber Transactions 26
- 27. Waiting in a Queue is Not Bad www.sharepointsummit.org Azure Queues Storage Account Queue Submitqueue my-account Calc-queue Message Joe’s Job Mary’s Job Andy’s Job 27
- 29. Acting Asynchronously www.sharepointsummit.org We can optimize the server throughput with async calls to the next tier Web Tier Web Tier Request 1 Response 1 Request 2 Request 1 Thread #1 Data Tier Response 1 Thread #1 Request 2 Just got easier in NET 4.5 with async/await Data Tier
- 31. Redundant Design www.sharepointsummit.org If any single node goes down, your app must continue unhindered Although some delay is OK Idempotent operations Load-balancing 31
- 32. Redundant Design www.sharepointsummit.org Azure Traffic Manager 32
- 35. 7 Key Things for Building a HighlyScalable SharePoint App www.sharepointsummit.org 1. Caching 2. Content Delivery Network / BLOB Storage 3. NoSQL 4. Sharding 5. Queuing 6. Asynchronous Processing 7. Redundant Design
- 36. Additional Resources www.sharepointsummit.org High Scalability http://highscalability.com/ Advanced Patterns in Cloud-Hosted Apps for Microsoft SharePoint http://channel9.msdn.com/Events/TechEd/NorthAme rica/2013/SES-B301 36
- 37. Thank you for your attention! This presentation will be available on the Vancouver SharePoint Summit web site a few days after the event.
- 38. Please rate this session! Fill out the survey and get a chance to win a Surface