1400-Sensor Cloud-NTU(Lim Hock Beng)

http://http://www.ntu.edu.sg/intellisyswww.ntu.edu.sg/intellisys 11

Sensor Cloud : Towards Sensor Cloud : Towards SensorSensor --Enabled Cloud ServicesEnabled Cloud Services

Dr Lim Hock BengIntelligent Systems Center

Nanyang Technological University13 Apr 2009


� Cloud Computing� A Scenario� Concept of Sensor Cloud� Research Challenges� Sensor Cloud Architecture� Key Components� Sensor Data Sources

OverviewOverview


� A new buzzword with intricate connections to Grid Computing, Utility Computing, Cluster Computing and Web 2.0 which are well established.

� Many attempts to define it . . .� Cloud Computing refers to both the applications delivered as services

over the Internet and the hardware and systems software in the data centers that provide those services.

� An infrastructure in which we won’t compute on local computers, but on centralized facilities operated by third-party compute and storage utilities.

� A large-scale distributed computing paradigm that is driven by economies of scale, in which a pool of abstracted, virtualized, dynamically-scalable, managed computing power, storage, platforms, and services are delivered on demand to external customers over the Internet.

� A way to increase capacity or add capabilities on the fly without investing in new infrastructure, training new personnel, or licensing new software.

Cloud ComputingCloud Computing


Broadly, It Looks Like Broadly, It Looks Like ……

Social Social

NetworksNetworks

App App

ServersServers

Mobile Mobile

ComputingComputing

CodeCode

InstrumentsInstruments

Data StorageData Storage

PlatformsPlatforms

Applications / ServicesApplications / Services

EconomicsEconomics SecuritySecurity


On the Architectural LevelOn the Architectural Level

Programming ModelsProgramming Models

SLA and SLA and

QoSQoS

NetworksNetworksStorage TechnologiesStorage TechnologiesOperating Systems Operating Systems

and Environmentsand Environments

Virtual StorageVirtual StorageVirtual Services Virtual Services Virtual ServersVirtual ServersVirtual Virtual

NetworksNetworks

MonitoringMonitoringSchedulingSchedulingDynamic Dynamic

ProvisioningProvisioningInformation Information

ManagementManagement

BioinformaticsBioinformaticsSocial Social

NetworkingNetworkingScientific Scientific

SimulationsSimulationsFinanceFinance

Environ. Environ.

MonitoringMonitoringEnterprise Enterprise

ComputingComputing

System System

InfrastructuresInfrastructures

Hardware & PlatformsHardware & Platforms

Live RelocationLive Relocation

VirtualizationVirtualization

RobustRobust

ResourceResource

AllocationAllocation

ManagementManagement

ApplicationsApplications

Security & Scalability

Security & Scalability

Economics of Cloud

Economics of Cloud


� Immense computational and storage resources that are collocated

� Very high speed data processing and movement� Accessibility over the Internet� Service-oriented Architecture� Accessibility from virtually any platform and device

Key Features of our InterestKey Features of our Interest


� The immense power of the Cloud can only be fully exploited if it is seamlessly integrated into our physical lives.

� That means – providing the real world’s information to the Cloud in real time and getting the Cloud to act and serve us instantly.

� That is – adding the sensing capability to the Cloud.

Cloud is Limited Cloud is Limited –– as of nowas of now


� Seamlessly couples the physical environment with the digital world.� Sensor nodes are small, low power, low cost, and provide multiple

functionalities� Sensing capability, processing power, memory, communication

bandwidth, battery power.� In aggregate, sensor nodes have substantial data acquisition and

processing capability.� Useful in many application domains – Environment, Healthcare,

Education, Defense, Manufacturing, Smart Home, etc.

Wireless Sensor Networks (Wireless Sensor Networks ( WSNsWSNs))

MICA mote sensor boards Telos mote Speck


� Very challenging to scale sensor networks to large sizes.� Proprietary vendor-specific designs. Difficult for different

sensor networks to be interconnected.� Operate in separate silos. Sensor data cannot be easily

shared by different groups of users.� Insufficient computational and storage resources to

handle large-scale applications.� Used for fixed and specific applications that cannot be

easily changed once deployed.

� Slow adoption of large-scale sensor network applications.

Sensor Networks are Limited tooSensor Networks are Limited too


The Missing PieceThe Missing Piece

Sensor NetworkSensor Network


A ScenarioA Scenario“Lets go to the Northern Cliffs”

““Lets go to the Northern Lets go to the Northern

CliffsCliffs””

“Sounds good. 1: Please take your lunch as you appear hungry.2: Carry drinking water as water at the Cliffs is contaminated.3: Use anti-UV skin cream.”

““Sounds good. Sounds good.

1: Please take your lunch as 1: Please take your lunch as

you appear hungry.you appear hungry.

2: Carry drinking water as 2: Carry drinking water as water at the Cliffs is water at the Cliffs is contaminated.contaminated.3: Use anti3: Use anti--UV skin cream.UV skin cream.””

1:1: 2:2:

3: Map to nearest food outlets3: Map to nearest food outlets

4: Take pictures of restaurant and send images4: Take pictures of restaurant and send images

5: Menus of restaurants and recommended food5: Menus of restaurants and recommended food

6: 6: ““Your Your FacebookFacebook

friends is dining near by friends is dining near by

Go catch up with her !Go catch up with her !””77


� Cell phone records the tourist’s gestures and activates applications such as camera, microphone, etc.

� The cell phone produces very swift responses in real time after:� Processing geographical data� Acquiring tourist’s physiological data from wearable physiological

sensors (blood sugar, precipitation, etc) and cross-comparing it with his medical records

� Speech recognition� Image processing of restaurant’s logos and accessing their

internet-based profiles� Accessing tourist’s social network profiles to find out his friends

� Fact : the cell phone cannot perform so much tasks !

An Insight into the ScenarioAn Insight into the Scenario


� Acquisition of data feeds from numerous body area (blood sugar, heat, perspiration, etc) and wide area (water quality, weather monitoring, etc) sensor networks in real time.

� Real-time processing of heterogeneous data sources in order to make critical decisions.

� Automatic formation of workflows and invocation of services on the cloud one after another to carry out complex tasks.

� Highly swift data processing using the immense processing power of the cloud to provide quick response to the user.

Need to Integrate Sensors with CloudNeed to Integrate Sensors with Cloud


An infrastructure that allows truly pervasive computation using sensors as interface between physical and cyber worlds, the data-compute clusters as

the cyber backbone and the internet as the communication medium.

� It integrates large-scale sensor networks with sensing applications and cloud computing infrastructures.

� It collects and processes data from various sensor networks.

� Enables large-scale data sharing and collaborations among users and applications on the cloud.

� Delivers cloud services via sensor-rich mobile devices.� Allows cross-disciplinary applications that span

organizational boundaries.

The Sensor CloudThe Sensor Cloud


� Enables users to easily collect, access, process, visualize, archive, share and search large amounts of sensor data from different applications.

� Supports complete sensor data life cycle from data collection to the backend decision support system.

� Vast amount of sensor data can be processed, analyzed, and stored using computational and storage resources of the cloud.

� Allows sharing of sensor resources by different users and applications under flexible usage scenarios.

� Enables sensor devices to handle specialized processing tasks.

The Sensor CloudThe Sensor Cloud


� Complex Event Processing and Management� Real-time data feeds from heterogeneous sensors trigger certain events

and services.� The data can be used by the applications to identify the contexts and the

locations and environment where the devices are being used. � Find out what other services and data may also be relevant to the current

set of data in order to make the correct decisions.� Massive Scale and Real Time Data Processing

� Integration with heterogeneous and massive data sources is a challenge due to the amount of information to be mined and used in real time.

� If the data includes real-time multimedia content such as streaming video, voice and images, it is a challenge to accurately process the data.

� It is also challenging to classify such content to trigger relevant services that may assist the user in his current location and scenario.

Research ChallengesResearch Challenges


� Large Scale Computing Frameworks� Multiple sensor data sets used for decision making may or may not be

collocated. � If these data sets and their corresponding access/search services are

geographically distributed, the allocation of computational and storage and data migration become critical challenges.

� Harvesting Collective Intelligence� Heterogeneous and real-time sensor data feeds allow us to improve the

decision making by using data and decision level fusion techniques. � To maximize the intelligence that can be exploited from massively

collocated information in a cloud is a challenge.

Research ChallengesResearch Challenges


Sensor Cloud ArchitectureSensor Cloud Architecture

Sensor Network Sensor Network

ProxyProxy

Satellite ConnectionSatellite Connection

I N T E R N E TI N T E R N E T

SensorSensor--Cloud Cloud

ProxyProxy

Traditional Traditional

CloudCloud

UserUser

Standalone GPS Standalone GPS

SensorSensor

Network of Sensors with Network of Sensors with

Internet ConnectivityInternet Connectivity

Handheld with Handheld with

SensorsSensors


� Sensor-Cloud Proxy� Interface between sensor resources and the cloud fabric.� Manages sensor network connectivity between the sensor

resources and the cloud.� Exposes sensor resources as cloud services.� Manages sensor resources via indexing services.� Uses cloud discovery services for resource tracking.� Manages sensing jobs for programmable sensor networks.� Manages data from sensor networks

• Data format conversion into standard formats (e.g. XML)• Data cleaning and aggregation to improve data quality• Data transfer to cloud storage

� Sensor-cloud proxy can be virtualized and lives on the cloud !

Key ComponentsKey Components


� Sensor Network Proxy� For sensor resources that do not have direct connection to the

cloud, this component provides the connection.� The sensor network is still managed from the Sensor-Cloud

Interface via Sensor Network Proxy.� The proxy collects data from the sensor network continuously or

as and when requested by the cloud services.� Enhances the scalability of the Sensor Cloud.� Provides various services for the underlying sensor resources,

e.g. power management, security, availability, QoS.

Key ComponentsKey Components


� Develop a sensor cloud prototype based on the Open Cirrus cloud computing testbed in Singapore.

� Use existing sensor deployments in Asia-Pacific as real-time information sources.

Sensor Cloud PrototypeSensor Cloud Prototype


� Large-scale environmental study project in Singapore� Supported by multiple government agencies (NEA, MOE, etc)� Promote awareness about weather patterns, climate change,

global warming and the environment among Singapore’s youth.� Mini weather stations deployed in schools throughout Singapore.

� Goals of the National Weather Sensor Grid (NWSG)� Connects the mini weather stations deployed in all schools.� Collects and aggregates weather data into a Central Data

Depository in a continuous, pervasive, and real-time manner.� Process, analyze, and store large amounts of weather data using

grid resources.� Provide geo-centric web interfaces to access the data.� Weather data can be accessed and shared by multiple schools,

user agencies, and the public.

Sensor Data Sources Sensor Data Sources ––The National Weather Study ProjectThe National Weather Study Project


� Earth Observatory of Singapore (EOS) - A new research initiative using sensor-rich infrastructure to provide: � Efficient monitoring.� Early alarming.� Decision making for critical events such as earthquakes, volcanic

eruptions, tsunamis, etc.� Under EOS, the Sumatran cGPS Array (SuGAr) is a

continuous GPS network formed by 29 GPS stations deployed along the Sumatran Plate boundary.

� It serves as the main GPS data source for The Sumatran Plate Boundary Project, which is a multi-disciplinary effort to understand tectonic processes at a plate boundary.

� We are developing a sensor grid for collecting data from the GPS stations, and to process, visualize and manage the GPS data.

Sensor Data Sources Sensor Data Sources ––Sumatran Sumatran cGPScGPS Array (Array ( SuGArSuGAr ))


� An initiative that aims to encourage the development of technologies for disaster/emergency detection, mitigation, response, and recovery in the Asia-Pacific region.

� Collect data from environmental sensors deployed in the participating APEC countries to form knowledge base.

� Hundreds of weather sensors as part of LiveE! (Japan) and NWSG (Singapore) projects have been deployed in Japan, Singapore, Taiwan, China, Thailand, etc.

� The weather sensors used are heterogeneous and multi-vendor.

� We are developing mechanisms for seamless integration and sharing of data and knowledge.

Sensor Data Sources Sensor Data Sources ––AsiaAsia --Pacific Environmental Sensor GridPacific Environmental Sensor Grid


Questions ?Questions ?


We are interested to discuss about the use of cloud computing for real-world applications, and explore opportunities for collaboration. Please contact :

Dr Lim Hock BengIntelligent Systems CentreNanyang Technological UniversityPhone : +65-6514-1005Email : [email protected]

ContactContact

1400-Sensor Cloud-NTU(Lim Hock Beng)

Documents

Transcript of 1400-Sensor Cloud-NTU(Lim Hock Beng)