Project Flogo

Open Source framework for IoT Integration. See flogo.io.

Transcript

1. Open Source Framework for IoT Integration Rajeev Kozhikkattuthodi TIBCO Software @rajeevtk
2. Open Source Framework for IoT Integration Ultralight to run on IoT devices Web UI for Design & Auto-Stepback Debugging Edge-native design for resilience
3. Agenda •  IoT Integration challenges •  Designing IoT Integration •  What’s Project Flogo? •  What’s next? •  Q&A
4. IOT INTEGRATION What’s the big deal with it anyways?
5. Extreme Scale 21 billion connected devices by 2020 Extreme Fragmenta/on 125+ IoT Pla7orms Digital-Na/ve UX 70% users are Millennials & Gen-Z by ‘20 Looming Security Risk Immature Security PracDces The Perfect Storm in IoT Image credits : NaDonal Ocanic and Atmospheric AdministraDon/Wikipedia Business Models in flux Early stage market risk SoBware eats the world Unclear ecosystem dynamics
6. Gartner Market Guide for IoT Integration 2016 “Through 2018, half the cost of implementing IoT solutions will be spent on integration.” Vodafone 2016 IoT Barometer Report “IT integration is key to using IoT data effectively. IoT is all about data, but 81% of business said that IoT can only deliver real value if effective use is made of the data generated” http://www.vodafone.com/business/iot/the-iot-barometer-2016 There is no IoT without Integration
7. IoT Landscape 2016 Edition Matt Turck & David Rogg, First Mark Capital 125+ IoT platforms
8. What’s an Enterprise got to do? Executive Sponsorship Rapidly evolving Business Models Software as a core competency Establishing Business Case Proprietary lock-in API-Only integration approaches Software bloat on edge devices Poor Developer & DevOps Tooling Fragmented platforms Unreliable connectivity Lopsided cost models New security practices Ecosystem Dynamics Total Cost of Ownership Flexible Pricing Models Smart-cloud but Dumb-Edge
9. INTEGRATING IOT So let’s dig in..
10. IoT Integration Myth #1 IoT Integration = APIs Our Point of View Yes, API-1st, but more.
11. Event-driven Microservices •  Model events, not just APIs •  Domain-driven design principles apply •  Examples of events – Temperature, Pressure, Humidity, Light, etc. – Device Activation, Reboot/reset, Timer, etc.
12. IoT Integration Myth #2 App logic runs in the cloud, only data collection runs in the edge Our Point of View Build Edge apps on the device to complement the cloud.
13. Typical Cloud-only IoT Integration scenario What’s wrong? •  Connectivity is unreliable •  Non-trivial latency in hops •  Connectivity increases TCO
14. A better IoT Integration blueprint Edge integration & event processing •  Local control = more reliable •  Less bandwidth & lower TCO •  Fewer hops & latency
15. IoT Integration Myth #4 We cant afford loosely coupled microservices on edge devices. Our Point of View Yes, you can. And you must.
16. Running monolithic edge IoT apps & expecting to scale
17. Examples of Edge Microservices
18. IoT Integration Myth #3 Build for the cloud, run on the edge device. Our Point of View Build Edge-Native, not just Cloud-Native.
19. On-Prem Infra “My Pet” Cloud Infra “Herd of Cattle” IoT Edge Infra “Swarm of Bees” © User:Mark Osgatharp / Wikimedia Commons / CC-BY-SA-3.0 © Keith Weller, USDA ARS / Wikimedia Commons / CC-BY-SA-3.0 © JimmyDominico, Pixabay / CC0 Public Domain
20. Cloud Infrastructure “Herd of cattle” Infinite compute scale Elastic resource pool Reliable connectivity Reliable shared storage Optimized for general purpose processing Identity & location agnostic Edge Infrastructure “Swarm of bees” Finite compute scale Limited resources Unreliable connectivity Unreliable local storage Optimized for I/O processing Identity & location matters Cloud-NaDve Edge-NaDve
21. Meditation on inevitable death should be performed daily. Every day when one's body and mind are at peace, one should meditate upon being ripped apart by arrows, rifles, spears and swords, being carried away by surging waves, being thrown into the midst of a great fire, being struck by lightning, being shaken to death by a great earthquake, falling from thousand-foot cliffs, dying of disease or committing seppuku at the death of one's master. - Yamamoto Tsunetomo Inspiration for Anti-Fragile Edge Apps
22. Edge-Native Application Checklist •  Externalize any application-state •  Think events, not just APIs or services •  Think async messaging, not just Request-Response •  Model “digital shadows/twins” in the cloud •  Privilege autonomous operations
23. Key takeaway Build event-driven edge microservices. Design them the edge-native way.
24. WHAT IS FLOGO?
25. Moonshot mission •  10X lighter engine •  10X productive user experience •  Edge-native design
26. Open Source Framework for IoT Integration Ultralight to run on IoT devices Web UI for Design & Auto-Stepback Debugging Edge-native design for resilience
27. Hello, Flynn! A hummingbird that epitomizes Project Flogo’s values of lightness, agility and fun.
28. The lightest integration. Ever. 50X lighter than Java? 20x lighter than Node?
29. Really?? More than 10x lighter? Yes. Custom-built “fit for purpose” engines per app with zero overhead.
30. Comparing Flogo with alternatives Eclipse Kura Node-RED Flogo VM JVM V8 Not applicable Base Disk/ Container 59 MB 56 MB ~ 0 MB (Scratch container) Base Runtime ~ 170 MB (with Open JDK) ~ 20 MB (with Oracle Embedded Java) > 50 MB ~ 5 MB Startup time Slow (~8 sec) Slow (~5 sec) Fast (~1 sec) Application Build Slow Not Applicable Fast (~2 sec) Based on early testing & publicly available information
31. Developing integration apps in Flogo 1 Design & Debug flows in Web UI 2 Package into apps using CLI or CI/CD 3 Run directly on edge gateways & devices
32. Web UI to design flows •  Low friction web-native UX •  Express app logic using rich flows, not just data or request pipelines •  Inline data transformations •  Build against a variety of IoT device platforms – ARM, x86, etc.
33. Flogo End User Concepts Flogo App = Flows + Triggers •  Triggers start flows •  Flow consists of Activities & Transitions •  Optional flow-level error handler Trigger Flow AcDvity Error Handler
34. Flogo – what’s going on behind the scenes Trigger(s) AcDon AcDvity AcDvity AcDvity Flow Your custom Go code github.com/TIBCOSojware/flogo-lib/core/acDon Custom Ac/on App Flow Service (for Design & Debug) State Service
35. State of the Art in IoT Debugging c. 2015
36. Web-based Auto-Stepback Debugger •  Debug edge apps in web browser •  Simulate external triggers like sensor or gateway events •  Automagically step-back in a flow to observe output for any activity •  Replay from trigger or any activity
37. We chose Golang •  Lightweight & efficient •  Strict typing •  Fast build times •  Zero-dependency binary executables •  Optimized for multi-core •  Awesome for large code bases •  Great community!
38. Zero-dependency app model Hardware OS VM - JVM Framework - OSGi App Java based Eclipse Kura Hardware OS VM – V8 Framework - Node App NodeJS based NodeRed Hardware OS App Golang-based Project Flogo Typically 100s of dependencies – 100MB+ Zero-dependency apps enabled by Golang
39. Hardware OS Security Logging Metrics Protocol Bridges DiagnosDcs Sensor Processing Make room for what matters – your apps Flogo Hardware OS VM - JVM Framework - OSGi App Java based Eclipse Kura Hardware OS VM – V8 Framework - Node App NodeJS based NodeRed Zero overhead means you don’t run out of room for real apps
40. Zero-dependency model •  Self-contained binaries for target platforms •  Deterministic build & operational behavior •  No dependency hell to manage – No messing with Java Classpath & Classloaders – No unwieldy package.json dependencies •  Easy to plug into your DevOps pipeline of choice
41. Edge-native Design Externalize state of apps on unreliable devices Examples •  Resilience: Device activation app that survives device cold-boots •  Debugging: Remote forensics & debugging of faults in apps Trigger(s) AcDon Flow Service (Design/Debug only) State Service Pluggable transports like HTTP, MQTT, etc. App
42. Beware the class hoarder! Flogo builds custom engines that bring along none of the stuff you don’t need. © Jeff Werner Flickr / by-nc-sa hqps://www.flickr.com/photos/jeffwerner/2308087644
43. Custom-built engines Users retain full control over activities and triggers included in an engine •  Zero waste – no class hoarding like Java •  Improved security – include only what you trust •  Reject “out of policy” applications
44. Intel IoT Architecture PotenDal Flogo™ use cases •  Event-driven flows •  Device Lifecycle Mgmt •  User Experience FTL/eFTL BW Container EdiDon Microflow CE Mashery
45. Amazon IoT Reference Architecture Flogo Cloud use cases •  Device Lifecycle Mgmt •  OrchestraDon BusinessWorks Microflows Mashery Flogo Edge use cases   Event-driven Flows   Device Lifecycle Mgmt
46. Project Flogo and Google IoT Architecture
47. WHAT’S NEXT FOR FLOGO?
48. What’s next for Flogo? Lighter •  Run Flogo nanoservices on microcontrollers Simpler •  ApplicaDon modeling in Web UI •  ConversaDonal Bot design interface Beqer •  Support for 12FA & container tooling in PaaS •  More connectors & extensions to IoT & SaaS endpoints
49. Flogo Nanoservices on Microcontrollers Flogo nanoservices that target microcontrollers powered by ARM M0 & higher Offload I/O centric compute farther off to the edge. So what? •  Event processing logic on microcontrollers
50. Flogo ApplicaDons Flogo Device Adapters (Developer Preview in November) TIBCO Cloud IntegraDon