Sensor Grid: Integration of Wireless Sensor Networks and the Grid Authors: Hock Beng Lim, Yong Meng Teo, Protik Mukherjee, Vihn The Lam, Weng Fai Wong,

Sensor Grid: Integration of Wireless Sensor Networks and the Grid

Authors: Hock Beng Lim, Yong Meng Teo, Protik Mukherjee, Vihn The Lam, Weng Fai Wong, and Simon See

Presentation: Maria Vanina Martinez

Wireless Sensor Networks

• WSNs can be seen as platforms with the potential to couple the digital world to the physical world.

• They are possible due to the development of new technologies such as MEMS sensor devices, wireless networking, and lower-power embedded processing.

• WSNs are composed by small, low-cost, low-power and self-contained devices that have the capability to sense, process data, and communicate via wireless connections.

WSN Applications

• Applications require interaction between the user and the physical environment.

• WSN applications include environmental and habitat monitoring, healthcare, military survelliance, tracking of goods, etc.

• Each sensor has limited capabilities, but when a large number is deployed and aggregated over a wide area, WSNs become important distributed computing resources.

Grid Computing

• Grid computing is an approach for the coordinated sharing of distributed and heterogeneous resources.

• It seeks to solve large-scale problems in dynamic virtual organizations.

• There exist many kinds of grids, but most of the existing developments are based on data and computation grids.

• Examples: SETI@home, GIMPS, etc.

Rationale for Sensor Grids

• All the data collected by sensors (it can be a lot) can be processed, analyzed, and stored using the grid’s resources.

• It is possible for different users and applications to flexibly share sensors.

• There are computationally powerful sensor devices, so it is more efficient to off-load specialized tasks to sensor devices (i.e. image and signal processing)

• Sensor Grids provide seamless access to a wide variety of resources in a pervasive manner.

Rationale (Cont.)

• Advanced techniques in AI, data fusion, data mining, and distributed database processing can be used to:

– make sense of all the collected data

– generate new knowledge about the environment

• Results can be used to:

– optimize the operation of the sensors

– influence the operation of actuators to change the environment

The Paper’s Contribution• The paper proposes a Sensor Grid architecture:

Scalable Proxy-based aRchItecture for seNsor Grid (SPRING).

• The main idea is to use proxy systems as interfaces between the WSN and the grid fabric.

• The authors present a series of challanges and design issues, addressing them while describing the architecture.

• They developed a sensor grid testbed to study the design issues and improve the architecture.

Design Issues and Challenges

• Grid APIs for Sensors

• Network Connectivity and Protocols

• Scalability

• Power Managment

• Scheduling

• Security

• Availability

• Quality of Service

Grid APIs for Sensors• Adopt grid standards and APIs for integration.

• The Open Grid Service Architecture (OGSA) is based on standards and technologies like XML, SOAP, and WSDL.

• If sensor data were available in the OGSA framework, it would be easier to exchange and process data on the grid.

• It is not possible to encode the data in XML format within SOAP envelopes in sensors.

• Grid services are too complex to be implemented on sensors.

Network Connectivity and Protocols• Network connections in grids are usually fast and

reliable.

• The network connectivity in WSN is dynamic, and it might be intermitent and susceptible to faults (noise, degradation).

• Grid networking protocols are based on standard Internet protocols (TCP/IP, HTTP, FTP, etc).

• WSN are based on proprietary protocols (MAC protocol and routing protocols).

• Efficient techniques to interface both kinds of protocols are needed.

Scalability• The Sensor Grid should allow the easy integration of

multiple WSNs with grid resources.

• These WSNs may be owned by different virtual organizations (VO).

• Enable applications to access sensor resources across increasing number of heterogeneous WSNs.

Power Managment• Applications running on sensors must trade off

between sensor operation and battery life.

• Sensor nodes should provide adaptive power management facilities that can be accessed by applications.

• In the Sensor Grid, the availability of sensors does not depend only on their load, but also on their power consumption.

• The Sensor Grid’s resource management component has to take this into account.

Scheduling

• Scheduling of nodes in WSNs facilitates power and sensor resources management.

• A scheduler is needed in Sensor Grids for an efficient use of sensor resources by applications that collect data.

• Applications and users may submit many different kinds of jobs.

• The Scheduler must manage them in very different ways, since they may have different requirements.

Security• Organizations may share resources only if the

process is guaranteed to be secure.

• There are various proposals for security on Grids, such as Grid Security Infrastructure (GSI), the Security Assertion Markup Language (SAML), etc.

• WSNs are prone to security problems.

• Techniques to address these problems are sensor node authentication, encryption of data, and secure MAC and routing protocols.

• Sensor Grids require that security techniques of both sides be integrated seamlessly and efficiently.

Availability• Applications running on sensor nodes are prone to

failure.

• If a node is running out of power, or has failing HW, it should be possible to migrate jobs to another node.

• If possible, it would be convenient to replicate services in order to preserve results.

• The system should be able to recover and restart the interrupted jobs if unexpected interruptions occur.

Quality of Service

• Quality of Service determines whether a sensor grid can provide sensor resources on demand and efficiently.

• The QoS requirements of sensor applications must be described in a high-level manner.

• High-level requirements should be mapped into low-level QoS parameters that specify the amount of resources to be allocated.

• Service descriptions are also needed to express what the service does, how to access it, and the QoS parameters of the service.

Quality of Service (Cont.)

• It is also necessary to consider resource reservation, changes in resource availability, in network topology, and in network bandwidth and latency.

• Mechanisms to enforce QoS have been developed separately for WSNs and grids.

• In Sensor Grids, the QoS should be enforced in a coordinated manner, integrating mechanisms from both parts.

Sensor Grid Organization

• A sensor Grid consists of WSNs and conventional grid resources such as computers, servers, and disk arrays for processing and storing sensor data.

• Resources are shared, and possibly owned, by several virtual organizations (VO).

• Users from various VOs may access the resources in the sensor grid, even if the resources are not owned by their VO.

• The following figure shows a Sensor Grid and its components.

Organization of a Sensor Grid

The SPRING Framework• The paper proposes a proxy-based approach for a

sensor grid architecture.

• It allows sensor devices to be made available on the grid in the same way that conventional grid services are provided.

• Sensor services are resource-constrained.

• The proxy can support various different WSN implementations, which provides interoperability.

• The following figure shows the SPRING Framework.

The SPRING Framework

SPRING Features• SPRING is a layered architecture approach.

• Each layer represents the main software components that are used to build a Sensor Grid.

• Each layer defines services that are accessible via APIs by the application or other layers.

• The Grid Interface layer supports a standard grid middleware (i.e. Globus Toolkit) that enables different types of resources to communicate over the grid network.

The SPRING Framework

SPRING Features – User Side• The User Access layer provides an interface that

enables the submission of applications for execution.

• The applications may consist of sensor jobs that execute over the WSN to collect data, or computational jobs to process the sensor data.

• Sensor jobs are not multitasking in nature, and require specific durations and time slots.

• The Grid Meta-scheduler layer is used to schedule and route jobs according to their required resources.

The SPRING Framework

SPRING Features – WSN Side

• The WNS Proxy acts as an interface between the WSNs and the grid.

• The proxy has several important functions:

– It exposes the sensor resources as conventional grid services, making them available for any application.

– It translates the sensor data from its native format to a suitable OGSA format, such as XML.

– It provides the interface between the sensor network protocols and the Internet protocols.

The WSN Proxy Functions (Cont.)

– It mitigates the effects of unexpected sensor network disconnections (buffering, caching, link management).

– New WSNs can be integrated to the sensor grid just by adding proxy systems.

– The WSN Proxy also provides other services to address power management, scheduling, security, availability, and QoS for the underlying WSNs.

SPRING Features – WSN Side

• The WSN Management layer provides an abstraction of the specific APIs and protocols to access and manage the heterogeneous sensor resources.

• It manages the configuration of sensor nodes and provides status information about them.

• It also accepts sensor job requests from the grid and invokes the specific commands to execute the jobs on the sensor nodes.

SPRING Features – WSN Side• The WSN Scheduler is the local resource scheduler

for the WSN:

– It implements the low-level scheduling algorithms for sensor power and resource management.

– It controls the scheduling of sensor jobs requested by the user.

– Considering the job parameters, it checks the resource availability and reserves them.

– It works jointly with other Proxy Components to provide services for availability and QoS.

Proxy Software Architecture

Proxy Components

• The Data Management component:

– Converts sensor data from its native format to a grid-friendly format.

– Performs data fusion and optimizations to improve the quality of the collected data.

– It supports several methods for transferring the sensor data to the user application, such as using GridFTP, or data streams.

• The Information Services component advertises the available sensor resources as grid services, following the OGSA standards.

Proxy Components (Cont.)

• The WSN Connectivity component provides services to interface the WSN protocols and the grid networking protocols:

– Buffers the transmission of sensor data, caches the routing information of sensor nodes, and manages the ad hoc sensor network links.

• The Power Management component:

– Keeps track of the power consumption of the sensor nodes.

– It works together with the WSN Scheduler to preserve power on the sensor nodes.

Proxy Components (Cont.)• The Security component implements OGSA-compliant

grid security technologies.

• The Availability component:

– Monitors the sensor nodes for failing HW or weak power levels, and migrates the jobs to more reliable nodes.

– It can replicate services and manage recovery of interrupted jobs.

• The QoS component, together with the Scheduler and the WSN Connectivity component:

– Performs the reservation and allocation of sensor resources based on QoS requirements of sensor jobs.

– It adapts networking conditions to provide the desired QoS.

The SPRING Framework

SPRING Features – Resource Side

• The Resource Management layer provides APIs to access and manages the resources for the grid job executions.

• These resources are distributed and heterogeneous computational and storages devices.

• The Resource Scheduler performs scheduling over grid jobs based on local usage policies.

Implementation• The authors developed a prototype sensor grid

testbed.

• They used the testbed to study the design issues using real hardware.

• They completely implemented the Grid Interface layer common to all the parts in the framework, and the layers from the user and resource sides.

• In the WSN Proxy they implemented the WSN Scheduler, and the WSN Management layer.

• Current work is being dedicated to implementing the Proxy components.

Conclusions

• The integration of wireless sensor networks with grid computing greatly enhances the potential of both technologies for new and powerful applications.

• Sensor grids will attract growing attention from both the research community and the industry.