FIRE! IoT-enabled Building 2018 ACM/IFIP International Middleware - - PowerPoint PPT Presentation

▶
fire
SMART_READER_LITE
LIVE PREVIEW

FIRE! IoT-enabled Building 2018 ACM/IFIP International Middleware - - PowerPoint PPT Presentation

Jan 27, 2024 274 likes •472 views

FireDeX: a Prioritized IoT Data Exchange Middleware for Emergency Response Georgios Bouloukakis 1,2 Rennes, France, December 2018 2018 ACM/IFIP International Middleware Conference Joint work with Kyle Benson 1 , Casey Grant 3 , Valrie Issarny 2

slide-1
SLIDE 1

FireDeX: a Prioritized IoT Data Exchange Middleware for Emergency Response

Georgios Bouloukakis1,2

Rennes, France, December 2018

2018 ACM/IFIP International Middleware Conference Joint work with Kyle Benson1, Casey Grant3, Valérie Issarny2, Sharad Mehrotra1, Ioannis Moscholios4, Nalini Venkatasubramanian1

1Donald Bren School of Information and Computer Sciences, UC Irvine, USA 2MiMove team, Inria Paris, France 3National Fire Protection Association, USA

  • 4Dept. of Informatics & Telecommunications, Univ. of Peloponnese, Greece
slide-2
SLIDE 2

2018 ACM/IFIP International Middleware Conference Rennes, France, December 2018

Motivation: IoT-enhanced structural fire response

  • 2

Gas Sensor Heat Sensor Motion Sensor

IoT-enabled Building

Camera

Building Occupants

FIRE!

slide-3
SLIDE 3

2018 ACM/IFIP International Middleware Conference Rennes, France, December 2018

Motivation: IoT-enhanced structural fire response

  • 3

Emergency Dispatch Fire Department Fire fighters (FFs) & Equipment Incident Commander’s (IC’s) Dashboard Incident Command Post sensors Building

  • ccupants

analytics

slide-4
SLIDE 4

2018 ACM/IFIP International Middleware Conference Rennes, France, December 2018

Motivation: IoT-enhanced structural fire response

  • 4

FireDeX

 Problem: how to enable the exchange of heterogeneous data by taking into account stakeholders’ information requirements and network conditions as a scenario evolves? Heterogeneous IoT sources Different groups of stakeholders Constrained network Analytics

data size, rates & format relevance urgency new IoT sources IC FFs Building Occupants failed components lossy channels

slide-5
SLIDE 5

2018 ACM/IFIP International Middleware Conference Rennes, France, December 2018

The FireDeX approach

  • 5

Data Exchange Broker Bandwidth allocation policies Event prioritization

FireDeX Programmable network infrastructure

Information requirements Prioritized events

  • FireDeX middleware configures the data exchange & network with prioritization and

bandwidth allocation policies based on:

  • information requirements
  • network resource constraints
  • Rely on SDN to bridge critical information requirements with network flows.
  • Model the performance of FireDeX across multiple layers using Queueing Theory.
  • Use the underpinning formal model for deriving novel algorithms that prioritize IoT events

and tune notification delivery/response times.

IoT sources Events Emergency responders & people Subscription with utility function

Goal: timely and reliable delivery of the most critical data to relevant subscribers despite challenging network conditions.

slide-6
SLIDE 6

2018 ACM/IFIP International Middleware Conference Rennes, France, December 2018

SDN-background

  • 6

Apps Virtual Switches Physical Switches

Control Plane Data Plane

  • Net. Apps

Centralized Global Network State SDN Controller FireDeX App … … Northbound API Southbound API

  • net. flows
slide-7
SLIDE 7

2018 ACM/IFIP International Middleware Conference Rennes, France, December 2018

Mapping info. reqs. to network state

  • 7

1 ... N - 1 %

Subscriptions Connections Network flows Priorities Drop rates

App / DeX concept FireDeX configurations Topics Subscribers’ view Network view

  • Network flows enable SDN infrastructure to differentiate subscriptions (e.g. by UDP/TCP

port number + IP addr).

slide-8
SLIDE 8

2018 ACM/IFIP International Middleware Conference Rennes, France, December 2018

FireDeX across layers

  • 8

BMS

publish <topic>

Pub/Sub Broker

Event Prioritization & Bandwidth allocation policies

Unmodified! Use any impl…

subscribe <topic, utility>

  • info. reqs.

SDN Controller Situational Awareness Apps

e.g., FF & IC Dashboard, civilian alerts

situational awareness Packet Drop

SDN “big switch”

< “smoke”, 100 > < “water_pressure”, 50 >

slide-9
SLIDE 9

2018 ACM/IFIP International Middleware Conference Rennes, France, December 2018

Modeling FireDeX using queuing theory

  • 9

DX

𝑐𝑙

managed network

𝑦𝑙

𝑐𝑙, 𝑡𝑗

𝜇𝑜𝑝𝑢𝑗𝑔𝑧

𝑦𝑙,𝑔𝑘

𝜇𝑗𝑜

𝑦𝑙, 𝑠

𝑘

𝜈𝑝𝑣𝑢

multiclass & priority

Ω Φ

𝑄𝑠𝑘 ∈ 𝑄𝑆𝑦𝑙 ↔ M/M/1 M/M/1

𝑐𝑙

𝜇𝑜𝑝𝑡𝑣𝑐

𝑐𝑙, 𝑐𝑗

𝜇𝑔𝑥𝑒

M/M/1

𝜈 𝜈 𝜈 𝜈

multiclass

𝜈

𝑞0, 𝑤𝑘

𝜇𝑞𝑣𝑐

𝑞𝑗, 𝑤𝑘

𝜇𝑞𝑣𝑐 𝑞0 𝑞𝑗

…..

𝑠0

𝜇𝑡𝑣𝑐

𝑠𝑘

𝜇𝑡𝑣𝑐 𝑡0 𝑡𝑗

….. … Subscription matching

𝑐𝑙

𝜇𝑗𝑜

unmanaged network

𝑦𝑗

𝑣𝑛

𝑠j Our new queueing model

slide-10
SLIDE 10

2018 ACM/IFIP International Middleware Conference Rennes, France, December 2018

Prioritization algorithm

  • 10

Maximum utility achievable for subscription rj Serialized packet size for topic vj notifications Rate of notifications (publications) for subscription rj

  • 1. Estimate the adjusted utility function per

subscription: information value per unit of bandwidth.

  • 2. Sort subscriptions.
  • 3. Group them into approximately equal-sized

network flows.

  • 4. Priorities assigned to approximately equal-

sized groups of network flows.

𝑠j

𝐵 =

𝑠0 𝑠1 𝑠2 𝑠3 𝑠4 𝑠5 𝐵0 𝐵1 𝐵2 𝐵3 𝐵4 𝐵5 𝑠3, 𝐵3 𝑠1, 𝐵1 𝑠5, 𝐵5

,

𝑠0, 𝐵0 𝑠2, 𝐵2 𝑠4, 𝐵4 𝑠3 𝑠1 𝑠5 𝑠0 𝑠2 𝑠4 𝑔0 𝑔1 𝑔2 𝑄𝑠0 𝑄𝑠1 𝑄𝑠2

slide-11
SLIDE 11

2018 ACM/IFIP International Middleware Conference Rennes, France, December 2018

Drop rate algorithms

  • 11

“Rho tolerance” enables keeping a buffer within the bandwidth (~0.1)

  • 1. Formulated as a convex optimization

problem.

  • Maximize overall utility as sum of all

subscriptions’ utilities.

  • Enabled by choice of logarithm for

utility function.

  • 2. 2nd constraint: queue stability condition.
  • Ensures allocated bandwidth within

that available.

Flat Linear Exponential Optimized

Drop rates for each network flow Mapped priority to network flow

slide-12
SLIDE 12

2018 ACM/IFIP International Middleware Conference Rennes, France, December 2018

Experimental setup

  • 12
  • We validate our theoretical model, evaluate the FireDeX approach and compare different

prioritization and dropping algorithms.

  • We use JINQS (Java Implementation of a Network-of-Queues Simulation) to build our

queueing network: an open source simulator for building queueing networks.

  • We have extended JINQS to implement our new multiclass priority queueing model.
slide-13
SLIDE 13

2018 ACM/IFIP International Middleware Conference Rennes, France, December 2018

Model validation: varying traffic loads

  • 13
  • Analytical model for lowest

priorities is slightly less accurate.

 Under-loaded  Saturated  Overloaded

slide-14
SLIDE 14

2018 ACM/IFIP International Middleware Conference Rennes, France, December 2018

Model validation: scalability

  • 14
slide-15
SLIDE 15

2018 ACM/IFIP International Middleware Conference Rennes, France, December 2018

FireDeX approach evaluation

  • 15
  • With an overloaded system,

switch buffers fill up and cause high delay / packet drops.

  • Our approach delivers more high

priority events than finite buffers

  • nly.
  • High priority events also delivered

quicker.

  • Addition of drop rate policy

smooths success rate while reducing end-to-end delay.

slide-16
SLIDE 16

2018 ACM/IFIP International Middleware Conference Rennes, France, December 2018

Algorithms comparison

  • 16

Prioritization algorithms:

  • Our bandwidth-aware greedy

strategy performs better than bandwidth-unaware version.

  • Both better than no prioritization.
  • But random priorities are worst: need

to set priorities correctly!

Drop rate algorithms

  • Convex optimization performs best in

comparison to linear, exponential and flat policies (drop rates by assigned priority).

  • Plot shows varying utilities of async

events vs. data telemetry: simpler policies perform closer to optimal for larger differences.

slide-17
SLIDE 17

2018 ACM/IFIP International Middleware Conference Rennes, France, December 2018

Conclusions & Next steps

  • 17
  • We introduce a middleware that

integrates application and network awareness.

  • Our application-aware prioritization

algorithm improves the value of exchanged information by 36% when compared with no prioritization.

  • Network-aware drop rate policies

improve this performance by 42%

  • ver priorities only and by 94% over

no prioritization. Queueing model:

  • Consider non-Poisson arrival and

service rates by using G/G/1 or G/D/1 queues. System:

  • Alternative utility functions.
  • Tuning the entire broker network.
  • Use our TIPPERS testbed and CFAST

simulator to further evaluate the FireDeX approach.

Conclusions Next steps

slide-18
SLIDE 18

2018 ACM/IFIP International Middleware Conference Rennes, France, December 2018

Thank you

  • 18