DevOps Fest 2020. Pavlo Repalo. Edge Computing: Appliance and Challanges
- 1. Edge Computing: Appliance and Challenges DevOps Fest 2020
- 2. What I will share with you today β What is Edge Computing? β Where to use in business ? β Why it's going to be next big thing? β What it means for DevOps? β What were the practical Challenges? β Our vision of the future β Q&A
- 3. What is Edge Computing? Main aspects β resources virtualization β close to data origin β low latency for real time apps <20ms β distributed computing resources β autonomous operation capabilities
- 4. Typical Device Edge Computing approach
- 5. What is Edge Computing? Base Resources virtualization β CPU β GPU β RAM β Network β Flash memory Periphery access management β GPIO β Southbound radio β PCIe β USB β Audio/Video β Power β Industrial Bus(-es) β Sensor ports https://www.pine64.org/rockpro64/
- 6. Where to use in business? Application areas and use cases β Telecommunications β Manufacturing β Automotive β Mining, Oil & Gas β Energy β Agriculture β SpaceTech https://www.gartner.com/smarterwithgartner/what-edge-computing-means-for-infra structure-and-operations-leaders/
- 7. Where to use in business? CASE 1: Video processing Video analysis & enhancements Why I need to send video stream to cloud to extract plate ID?
- 8. Where to use in business? CASE 2: Factory 4.0 Robotics & Manufacturing I need to communicate locally, resiliently and at lowest cost
- 9. Where to use in business? CASE 3: Software Deο¬ned Satellite Space Tech β Camera and sensors storing data to local storage on satellite β Diο¬erent data models processing the data and supply it to the own infrastructure on the ground
- 10. Why going to be the Next Big Thing? Key factors of business success and why itβs serious β Market needs to connect clouds and IoT β Powerful governance by Linux Foundation & LF Edge β Major investments from key tech players Trillions of sensors & controls All kind of clouds
- 11. You Missed The Cloud Revolution? Edge is 4X the Size & Will Hit 4 Trillion Economy* http://www.chetansharma.com/publications/edge-internet-economy/ Arpit Joshipura, General Manager LF Networking, Edge and IoT at Linux Foundation. ONS 2019, Antwerp Belgium
- 26. Edge Computing: Key DevOps speciο¬cs What to take care specially β Resources management (scaling and periphery control) β Autonomous payload execution (especially for the Device Edge) β Security aspects: Secure boot
- 27. Edge Computing: our challenges Practical aspects: AgTech Why Edge Computing but not pure Embedded approach? Solution dev process - need to know IoT & domain speciο¬cs QA process - traditional vs reality Support&Maintanace aspects (SW upgrade, debug, monitoring) Upcoming challenges: Connectivity-as-a-Service (eSIM) Proof of data origin CV integration for non-AI speciο¬c HW
- 28. Practical aspects: AgTech β Itβs NOT βEmbedded vs cloudβ itβs βEmbedded PLUS Cloudβ β CI/CD & QA automation possibilities β Applications scalability β No chip dependency β Allows 3rd party applications β Imagine itβs like IOS or Android vs legacy ο¬rmware phone Why Edge Computing but not pure Embedded approach?
- 29. Practical aspects: AgTech Solution development process β Need to know IoT and domain speciο¬cs β GPS is not precise, not stable and needs to be ο¬ltered β Want to have stable precise positioning (RTK)? Stable connectivity required β Sensors (self-) calibration β Environmental conditions: outdoor, UV, Vibrations, chemicals, thermal management
- 30. Practical aspects: AgTech QA process: Load testing Traditional vs Edge
- 31. Practical aspects: AgTech QA: Staging environment Traditional vs Edge:
- 32. Practical aspects : AgTech Support & Maintenance aspects β Low qualiο¬ed operational personnel - make it plug&play or disable it β Operating in the middle of nowhere β Low speed & unstable connectivity: donβt expect stable telemetry & console β Foolproof user design is a must β If mobile network used - who is managing account balance? Credit: NASA/JPL-Caltech/MSSS
- 34. Upcoming challenges: β Deο¬ne TPM to be used β Implement secure boot by utilizing Keylime project β Enable digital signature for upstreaming data https://keylime.dev/ Proof of data origin & secure boot using TPM
- 35. Upcoming challenges: Computer Vision integration for non-AI speciο¬c HW β Training dataset collection (another use case for EdgeGW ) β Linking container with data model to speciο¬c GPU β Integration with autopilot for route correction for rows tracking