1 of 8

Towards Tunable Distributed Data Management for IoT

Luís Ferreira

luis.m.ferreira@inesctec.pt

HasLab - INESCTEC & University of Minho

​

Supervised by:

Fábio Coelho and José Pereira �(fabio.a.coelho,jose.o.pereira)@inesctec.pt

​

5th April 2022

16th EuroSys Doctoral Workshop (EuroDw 22)

2 of 8

Problem

Cloud won’t be able to meet IoT applications’ requirements by itself

3 of 8

Cloud won’t be able to meet IoT applications requirements by itself

    • 79.4 zettabytes of data by 2025
    • Network bottlenecks
    • Privacy and latency concerns
    • Limited connectivity scenarios

4 of 8

Fog databases as the solution

    • New requirements posed by Fog nodes:
      • Bigger clusters
      • Higher churn
      • Low resource nodes

​

5 of 8

Vision

End Devices

Fog Nodes

(Layer 1)

...

Tasks

Load-balancing

Data replication

Data aggregation

Service elasticity

Cloud

Fog Nodes

(Layer N)

Horizontal axis (quantitative)

Depth axis (qualitative)

Privacy�enforcement

6 of 8

How do databases fit into the edge/fog environment?

    • Use cases for DBMSs at Fog nodes are not clear
    • No solutions to reliably evaluate databases in the Fog

7 of 8

First steps

    • Survey: use cases, existing solutions (work in progress)
    • Benchmarking/simulation tool
    • Decide where to go from there…

8 of 8

Summary:

    • Problem: Cloud won’t be able to meet IoT applications’ requirements by itself
    • Vision: elastic, load-balanced, privacy enhancing Fog DBMS
    • How and when to use databases to process IoT data in the fog?
    • Next steps:
      • Survey
      • Benchmarking solution

�