edge-to-cloud solutions - moodle.msengineering.ch
TRANSCRIPT
TSM-CloudSys
Edge-To-Cloud solutions
Nabil Abdennadher
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022
The context (IoT today)• Thousands of connected IoT
sensors deployed at a large scale
• Tens (even hundreds) of applications coming, from different needs/domains/users, are emerging.
• Current IoT platforms tend to centralise intelligence in the cloud.
2
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022 3
The context: Centralised Cloud platforms
• AWS IoT Core• Azure IoT hub• Cloud IoT Core (Google)
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022
The context: From Cloud to Edge-Cloud• For many IoT applications, submitted to constraints
the centralised vision is not sustainable.• Data production is exploding, making Cloud based
centralised IoT platforms, ill-equipped to cope withthe huge quantity of collected data.• The Centralised model is doomed to fail.• The need to make the edge smarter is
inescapable.• A new model, combining edge and cloud, is
required4
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022
The context: A Hybrid Edge-Cloud platform
5
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022
PLAN
6
• What is Edge Computing ?• Use-cases• Edge-Cloud solutions
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022
What is Edge Computing ?
Edge computing is:• a method of optimising applications
by taking some “portions” away fromcentral nodes to the other extreme(the "edge").
7
• a practice of processing data near the edge of the network,where the data is being generated, instead of processing in acentralised data-processing warehouse.
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022
Why Edge Computing ?
• Four objectives are behind Edge computing:• Unload the cloud.• Limit the traffic between IoT devices and the
cloud• Keep decision as close as possible to the IoT
devices• Enhance security
8
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022
Why Edge Computing ?• Is there a need for near-real-time action on data
generated by sensors?• Is the data generated too big to transfer to the cloud?• Is the internet link between the sensors/actuators and the
cloud unreliable ?• Is there a privacy/security issue with transferring or
processing the data in the cloud (public or private)?
If the answer is yes to one or more of these questions, then you need an intelligent edge.
9
https://media.sixsq.com/blog/case-for-edge-computing
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022 10
Edge Computing markethttps://www.alliedmarketresearch.com/edge-computing-market
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022
What are the Cons of Edge computing ?• Strike the balance between
• “keeping data at the edge” and “bringing it intoa central cloud”
• “sophisticated algorithms in the cloud” and“lightweight analytical processes” in the edge?
11
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022
Fog Computing
• Edge computing usually takes place directly on theedge device to which the sensors are attached.
• Fog computing moves the edge computingactivities to a set of edge devices that belong tothe same LAN
• In Fog Computing, computing activities arephysically more distant from the sensors andactuators.
12
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022
Self-adaptive ML based applications
13
Closed/Feedback Loop
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022
Self-adaptive ML based applications
14
https://thenewstack.io/azure-iot-edge-a-technology-primer/
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022
PLAN
15
• What is Edge Computing ?• Use-cases• Edge-Cloud solutions
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022
PrinciplePrinciple
16
Low traffic
Medium traffic Medium light
Low light
High traffic Full light
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022
Smart Lighting Solution - Edge application
17
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022 18
Smart Lighting Solution - Edge application
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022
• Automatically recognise:• lanes, roundabouts, crossings
• Automatically set bounding boxes for traffic capture
19
Smart Lighting Solution - Edge application
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022
Smart public lighting
20
Edge controlled Intelligence
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022
Noise management
• Edge device: A noise “blind”detector based on acoustic and AItechnologies• 2 micros• Camera: only used for learning
• Classifying vehicles (cars, truck,buses, motorcycles, electric cars)
• Raspberry based prototype withpython software (NN), deployed atRue Jura (Geneva)
21
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022
From Monday Feb. 3 (5:00) to Friday Feb. (21:00): 13’783 vehicles
22
Noise management
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022
From Monday April. 6 (5:00) to Friday 10 April. (21:00): 7’178 vehicles
23
Noise management
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022 24
Noise management
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022
• The intelligence is context-aware: 30 km zone, 1 lane, 2 lanes, 4 lanes, weather, etc.
• The “context” is related to spatial and temporal conditions.
• Remotely control theintelligence in case ofmisbehaving sensors orswitching from one contextto another (Continuousintegration)
25
Noise management
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022 26
Noise management
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022
• The intelligence is context-aware: 30 km zone, 1 lane, 2 lanes, 4 lanes, weather, etc.
• The “context” is related to spatial and temporal conditions.
• Remotely control theintelligence in case ofmisbehaving sensors orswitching from one contextto another (Continuousintegration)
27
Noise management
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022 28
Edge controlled Intelligence
Noise management
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022
Data integrity• IoT applications often use low quality and/or non institutional
sensors• Low energy sensors: Most of IoT devices are battery powered
and should operate in an energy saving way.• Limited computing resources: minimal processing and
storage capability• Low cost/quality sensors: To keep the prices low, vendors
often use low quality hardware.• Not secure: Not build with security in mind
• à IT security techniques cannot be used
29
And here is where trust comes in …
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022
Data integrity
30
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022
Data integrity
~550 LoRa noise sensors
31
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022
Data integrity
• Signature
• Noisy streets, residential neighborhoods, 30 kmzones, etc.
• Precision/Accuracy
• Difference between the measure we receive andthe value a good quality sensor would give.
• Packet Error Rate
• Availability32
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022
Use-case 4: Signature Trust Factor
33
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022
Smart Grid
34
Power Grid
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022
Smart Grid
35
Micro-grid infrastructure Power Grid
Micro-grid infrastructure
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022
Smart Grid
36
Connection to the power Grid
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022
Smart Grid• Digital self-adaptive smart grid platform:
• An IoT-Edge-Cloud infrastructure composed of:• Home appliances & smart meters• Edge devices (yellow, blue and green boxes)• Cloud infrastructure
• A ML based self-adaptive application
37
Initial data Learning 1
New deployment
Failure HouseholdLearning N
.
.
.
Prediction 1
Prediction N
.
.
.
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022
Smart Grid
38
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022
PLAN
• What is Edge Computing ?• Use-cases• Edge-To-Cloud solutions
39
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022 40
Deploying on the edge-to-cloud platforms
T-CloudSys | Serverless / Function-as-a-Service | Academic year 2021/2022
Anatomy of Edge-To-Cloud solutions
• IoT Infrastructure Management: An orchestration service thatcomposes, provisions (configures and deploys) and monitorsEdge/IoT networks.
• Edge Framework: enable edge application modules programmingand execution.
• Container Facilities: build a container (such as Docker) with atrained MLM
• Communication Hub: create a messaging infrastructure fornetworked components
• Storage Facilities: store training data and several MLM versions inCloud data warehouses.
• Machine Learning Facilities: build and train a MLM in the Cloud.
41