dSpace: Composable Abstractions for Smart Spaces

Presentation

• 3x IoT devices, 13.6 per person by 2023

• Complex to program smart spaces with greater variety of devices and use cases

  • Device-centric APIs

  • Monolithic implementations

  • if-then--that policies

• Modularity

  • What is the right modularity?

  • What is the minimal set of abstractions to achieve it?

• Modules

  • Digivice: model and driver

    • Digivices can have different device libraries (driver) programming lang. (driver) schema (modl)

      • Heterogeneity --> complexity

  • Digidata: model and driver

    • mount digidata T to digivice D

• Composition

  • Mount

    • Idea:

      • use a universal digivice to configure a device-specific digivice

      • compose digivices with mount primitive

      • digivices form control hierarchy

        • Program the intent upward / downward

  • Pipe

  • Yield

    • Allow multiple parents, but only one can have write-access and other(s) are yielded

    • Device moves across rooms, delegate access to third-party

• Policies

  • Delegation

    • Must preserve multi-tree topology to avoid loops

Summary:

• Composable degivice / data

  • First-class primitives: mount, pipe, yield

  • Rich policies: delegation, intent reconc

  • Digis run as microservices