Se ha denunciado esta presentación.
Utilizamos tu perfil de LinkedIn y tus datos de actividad para personalizar los anuncios y mostrarte publicidad más relevante. Puedes cambiar tus preferencias de publicidad en cualquier momento.

IoT Meets the Cloud: The Origins of Edge Computing

History of edge computing: IoT meets the cloud. Lecture delivered as part of Duke University Electrical and Computer Engineering / Computer Science Special Topics course on Edge Computing designed and developed by the instructor.

Libros relacionados

Gratis con una prueba de 30 días de Scribd

Ver todo
  • Inicia sesión para ver los comentarios

IoT Meets the Cloud: The Origins of Edge Computing

  1. 1. ECE 590/COMPSI 590 Special Topics: Edge Computing Lecture date: Wednesday August 29th, 2018 IoT Meets the Cloud: The Origins of Edge Computing
  2. 2. Last Class: Introduction to Edge • Edge computing Advantages: latency, bandwidth, privacy Different devices Different degree of application centralization • Research themes 2
  3. 3. Barcelona PoC Deployment 3A New Era for Cities with Fog Computing, Yannuzzii et al
  4. 4. Edge Computing at Chick-fil-A (1/2) 4 July 2018
  5. 5. Edge Computing at Chick-fil-A (2/2) 5
  6. 6. This Class • Path towards the edge: Cloud computing • Path towards the edge: Internet of Things • Modern multi-tier architectures 6
  7. 7. Edge: IoT Meets the Cloud 7
  8. 8. The Pendulum 8
  9. 9. The Cloud: Applications and Providers • Amazon Web Services, Microsoft Azure, Google Cloud, IBM Cloud • Virtual machines, of different grades • An endless, always updating list of specialized services 9
  10. 10. Cloud Centralization: AWS Example 10
  11. 11. Cloud Centralization: Microsoft Azure Example 11
  12. 12. The Cloud: Massive Operation (1/2) 12
  13. 13. The Cloud: Massive Operation (2/2) 13
  14. 14. The Cloud: Shared Substrate • Shared servers • Shared cores • Shared network 14
  15. 15. Cloud: Some of the Properties • Geographically centralized • Massive, scalable • Managed, physically secure • Shared • Cloud outages are uncommon • … but task latency variations are the norm • … 15
  16. 16. Side Note: Cloud as an Enabler of Vibrant Web Ecosystem • Spurred innovation • Perhaps, edge doing the same for the IoT? 16
  17. 17. Edge Precursors: CDNs (1/3) • Content Delivery Networks Akamai, AWS CloudFront, Fastly • Original “edge nodes” 17
  18. 18. Edge Precursors: CDNs (2/3) • Content Delivery Networks - static content replication • Fewer points than in edge computing settings  E.g., Akamai: ~200,000, AWS CloudFront: 100 POPs 18
  19. 19. Edge Precursors: CDNs (3/3) • Interesting new development (2017): using CDNs to customize web server responses via Lambda@Edge • Possible research project: extending CDN mechanisms to edge computing 19
  20. 20. Edge Precursors: Peer-to-Peer • P2P: Napster, Kazaa, Bitcoin • File sharing • Focused on decentralization mechanisms above all 20
  21. 21. Related Area, for Some Edge Research: Distributed Workloads on the Cloud • … and in multi-core systems • Homogenous substrates • Non-responsive operations 21 • Research projects on edge analytics need to be specific about the differences in their settings and traditional ones
  22. 22. Intellectual Heritage: Distributed Clouds 22
  23. 23. Edge: IoT Meets the Cloud 23
  24. 24. This Class • Path towards the edge: Cloud computing • Path towards the edge: Internet of Things • Modern multi-tier architectures 24
  25. 25. History: IoT • Devices  smart devices  connected devices Thanks, Moore’s Law! 25
  26. 26. For Example, Towards IoT: Evolution of a Smart Watch • CES 2016 26
  27. 27. Side Note: By Now, Modern Cars Are All Electronics 27
  28. 28. Side Note: Progress in the IoT is Limited by Energy Storage • No Moore’s Law for batteries 28
  29. 29. IoT Properties (1/3) • Tightly constrained design space Often specialized for the application Proliferation of protocols and vendor-specific solutions 29
  30. 30. IoT Properties (2/3) • Low computing capacity E.g., laptop: 2.4 GHz, Raspberry Pi: 1.2 GHz, Arduino Due: 0.084 GHz, Amazon Dash Button: 0.016 GHz • Minimized/reduced energy consumption • Difficult to secure • … 30
  31. 31. Edge Precursors: Mobile Ad Hoc and Sensor Networks • Focused on sensing 31
  32. 32. Sensor Network Example: ZebraNet • Early 2000s 32
  33. 33. Sensor Network Example: RoombaNet 33
  34. 34. Sensor Networks: Focused on Multi-hop Connectivity 34
  35. 35. Industry Approach: 3-Tier Architectures Instead of Multihopping • Sensors → gateway → cloud 35
  36. 36. Edge and Sensor Networks: Differences • Research projects on resource discovery and peer assistance in edge need to be explicit about the differences in their settings and traditional ones 36 • No consideration of the cloud • No multi-point decision-making
  37. 37. This Class • Path towards the edge: Cloud computing • Path towards the edge: Internet of Things • Modern and envisioned multi-tier architectures 37
  38. 38. Barcelona PoC Deployment 38A New Era for Cities with Fog Computing, Yannuzzii et al
  39. 39. Multi-tier Architectures 39
  40. 40. 40 Smart city fog deployment: buildings, neighborhoods, regions connected with each other
  41. 41. Example Use Case: Securing Air Travel • Multiple locations need to work together • Cameras important part of the system  1 Tb/day/camera • Immediate action needed • Applications deployed: risk scoring, vehicle capture, baggage capture Airport terminal provisioned with a hierarchy of fog nodes
  42. 42. Current Platforms: AWS Greengrass 42 Released June 2017
  43. 43. Edge Properties • Decision-making, actuation • Data manipulations and transformations • Heterogeneity • Hierarchy Cloud is involved in the system • … 43
  44. 44. Class Recap • Origins of the edge, on the cloud side: CDNs, P2P systems • Origins of the edge, on the IoT side Sensor networks • Properties of edge systems 44
  45. 45. ECE 590/COMPSI 590 Special Topics: Edge Computing Lecture date: Wednesday August 29th, 2018 IoT Meets the Cloud: The Origins of Edge Computing

×