Scaling IoT Solutions at Dyson

Transcript

1. Scaling IoT Solutions
2. About me! • Tech team lead at Dyson • Been working in IoT for 14 years • Passionate about agile software development
3. Dysons Connected products
5. 360 Eye enhanced through connectivity • Quickly set a schedule • Has it cleaned my house? • Get notified when your clean has started
7. Pure cool enhanced through connectivity • Quickly set a schedule • How clean is my house? • Get ready for coming home • How have things changed over time?
8. Where are we now? • 100,000s of connected users • Across 30 countries • Increasing rate of new connected users
9. What does the future hold?
10. Transitioning from Products to Services
11. Marrying Waterfall with Agile • What are the requirements at launch? • Where do you put the UI? • User story mapping • Firmware update always!
12. Challenges Launching globally • Already established in multiple regions • Service availability & speed • User experience • Leveraging local experience
13. Where should the Logic sit?
14. What are the options? • What Features have to be connected? • Local Logic • Remote logic • Always be able to update!
15. Implications of local logic • Robust and fault tolerant • Fast • Hard to change • Must consider backwards compatibility
16. Implications of remote logic • Can be slower • Harder to make fault tolerant • Easier to change • What will scale?
17. Where do you schedule?
18. The challenges of Real-time control
19. Real-time product control • What has to be real-time? • What can be queued? • Consistency is key • What has to be saved?
20. Server load from IoT Devices
21. IoT Load the good & bad • Predictable • Throttled by manufacturing • Time synchronized • Does not give up!
22. Taming the IoT load • Consider data rates • Unsynchronize & back off • Queueing • Blacklisting
23. Automation with IoT Devices
24. IoT Load Testing • Leveraging Cloud computing • Use same metrics as production • Monitor the clients • Not just steady state http://grafana.org/
25. What Level do we automate? • Acceptance testing for sub systems • Contract Testing between systems • Smoke testing for the whole • Automate deployments for all testing
26. What Should you monitor? • Everything all the time! • Monitor KPIs not just server health • Try to monitor in one system • Treat IoT devices like servers • Trend as well as alert http://grafana.org/
27. Remaining backwards Compatible
28. What has to be compatible? • Firmware in products on 1st install • iOS & Android Apps • Micro services & backend • Avoid specific behavior for different regions • Do not to tie releases together
29. Security in an IoT World
30. What approach do we take? • Try not to do anything special! • Embed security into the development process • Don’t just test the server • Automate where possible
31. Learning and data mining
32. How do you get at the data? • Billions of anonymized data points • Get it off production • Get all of the data into the same system • Ask specific questions and leverage machine learning
33. How do we use the data? • Understand better how our customers use their products • Our Innovation continues after we ship the products • Even more powerful data driven design • Enabled by firmware update!
34. To wrap up
35. What are we learning? • Ensure connectivity extends the product • Consider what is required at launch and where the logic goes • Automate testing and deployment as much as possible • Use the data to continuously improve the products
36. We are Hiring! Careers.dyson.com
37. Thank you 