Scale your signalR realtime web application
5 likes4,352 views
This document summarizes a presentation about scaling out SignalR applications. SignalR allows real-time web functionality by pushing messages to clients using techniques like websockets. To scale SignalR, developers must understand how to distribute hub instances across servers and measure traffic patterns. The presentation covered best practices like using groups and async calls. It also described pluggable backends like SQL Server and Redis that can be used to share state across servers. Finally, it gave examples of custom scale-out scenarios using techniques like server transitions and filtered message buses.
Scale your signalR realtime web application
- 1. SELA DEVELOPER PRACTICE December 15-19, 2013 Ran Wahle Scale-out your Real-time signalR © Copyright SELA software & Education Labs Ltd. | 14-18 Baruch Hirsch St Bnei Brak, 51202 Israel | www.selagroup.com
- 2. Agenda What is signalR Do-s and Don’t-s Load patterns Messurement Scaleout with backplane Custom scaleouts
- 3. What is signalR Pub-Sub Ability to push-notify clients Uses WebSocket | SSE | ForeverFrames | Longpolling (best feature available) WebSocket ServerSentEvents = Long-Polling ForeverFrames
- 4. Scaleout Web application Number of requests Memory, CPU, Disk Bandwidth
- 5. Scaleout signalR It’s a web application Many long requests No cache (It’s real-time) Concurrency Different transports
- 6. SignalR Core Architecture Publisher Message cache Message Bus Worker Worker Worker Client Client Client Client
- 7. Do-s Understand Hub instances lifecycle Use as many general messages as you can Use groups Use async when calling methods from Hub
- 8. Don’t-s Use session Keep instance members on Hubs Send large messages
- 9. Load patterns Server broadcast Server push User event driven High frequency real-time
- 10. Server broadcast Low rate broadcast Same message / payload to all users More clients doesn’t effect message bus traffic
- 11. Server push Low rate broadcast Unique payload per client More clients More message bus traffic The rate is controlled
- 12. User event driven Broadcast on client action One message to many clients More clients More message bus traffic Users control the rate
- 13. High frequency Real time Fixed High broadcast rate One message bus send One user More clients More traffic
- 15. BackPlane Web-application are stateless signalR isn’t Client Client Client Client Hub Hub Client Client Client Client Client Client Hub Client Client Hub
- 17. Plugabble backplanes SQL Server Redis Azure service bus
- 18. Custom scaleout scenario Common Server Specific server Filtered service bus Server transition
- 23. Server push Ran is connected to me Server1 Server2 Ran
- 24. Server push Server1 Server2 Ran, I need your laptop Ran Yaniv
- 26. Server push Server1 Ran, I need your laptop Server2 Ran Yaniv
- 27. Server push Server1 Server2 Ran, I need your laptop Ran Yaniv
- 30. Questions
- 31. What have we seen today? signalR performance considerations Messuring and testing tools Pluggable back planes Custom scaleout scenarios
- 32. Thank You http://blogs.Microsoft.co.il/ranw [email protected] © Copyright SELA software & Education Labs Ltd. | 14-18 Baruch Hirsch St Bnei Brak, 51202 Israel | www.selagroup.com