Topic 3: Sensor Networks and RFIDs

32
Topic 3: Sensor Networks and RFIDs Part 6 Instructor: Randall Berry e-mail: [email protected] Northwestern University P 491: Selected Topics in Information Techno

description

Northwestern University MITP 491: Selected Topics in Information Technology. Topic 3: Sensor Networks and RFIDs. Part 6 Instructor: Randall Berry e-mail: [email protected]. Networking Issues. MAC layer issues Topology control Routing. Routing. - PowerPoint PPT Presentation

Transcript of Topic 3: Sensor Networks and RFIDs

Page 1: Topic 3: Sensor Networks and RFIDs

Topic 3: Sensor Networks and RFIDs

Part 6

Instructor: Randall Berrye-mail: [email protected]

Northwestern UniversityMITP 491: Selected Topics in Information Technology

Page 2: Topic 3: Sensor Networks and RFIDs

Networking Issues

• MAC layer issues• Topology control• Routing

Page 3: Topic 3: Sensor Networks and RFIDs

Routing

• In multi-hop networks, a routing algorithm needed to specify how packets reach destinations.

Page 4: Topic 3: Sensor Networks and RFIDs

Routing Challenges

• Algorithm must be adaptive to topology changes.– Due to mobility, node failures, etc.

• Limiting routing overhead is required due to energy considerations.

• Nodes may also have limited memory for routing tables.

Page 5: Topic 3: Sensor Networks and RFIDs

Routing considerations

• Types of flows:– Unicast, multicast, broadcast, geocast.

• Type of topology:– Tree, mesh.

• Routing metrics:– Shortest paths, minimum energy, maximum

lifetime.

• Availability of location information.

Page 6: Topic 3: Sensor Networks and RFIDs

Unicast Routing(Mesh topology)

• Many approaches:– Flooding, Gossiping.– Proactive routing (Table-based).– Reactive routing (on-demand).– Geographic routing.

Page 7: Topic 3: Sensor Networks and RFIDs

Flooding

• Simplest type of routing• Send any new packet to all neighbors (except one

received from)– “Wireless broadcast advantage” can help.

• Usually some type of expiration for each packet.• Minimal memory requirements/control traffic.• But excessive number of transmissions to reach

destination.– Topology control can help here

Page 8: Topic 3: Sensor Networks and RFIDs

Flooding/Gossiping

• Flooding can be useful for broadcast data.– Still can result in excessive transmissions/collisions.

• Gossip algorithms attempt to reduce this – Two variations:

1. Randomly pick one node to forward to.2. Each node only forwards a new packet with some

probability p.

– Cost?

Page 9: Topic 3: Sensor Networks and RFIDs

Proactive/Reactive Routing

• Both require nodes to keep a routing table.

• Difference is when table is updated.– Proactive schemes attempt to always have an

up-to-date table.– Reactive schemes only update tables “on

demand” i.e. when traffic to send.– Trade-offs?

Page 10: Topic 3: Sensor Networks and RFIDs

Proactive/Reactive Routing

• The routing algorithms used in the internet are an example of proactive routing.– Periodically routers exchange distance vectors or link states.

– Every router has a routing table with up-to-date routes.

• An example of a reactive protocol is the AODV (Ad hoc On demand Distance Vector) protocol– routing option for Zigbee mesh networks.

• Various hybrids also exist.

Page 11: Topic 3: Sensor Networks and RFIDs

AODV

• In AODV nodes do not exchange any routing info.

• To establish a route to a new destination a node broadcasts a route request (RREQ) packet to all neighbors.

• RREQ contains destination address, source address, hop count, broadcast id, and source and destination sequence numbers.

Page 12: Topic 3: Sensor Networks and RFIDs

AODV

• When node receives a RREQ:– If already seen – does nothing.– If new and no route to destination - increments hop

count, forwards RREQ, and stores routing entry to source.

– If knows route to destination – then responds with a route reply (RREP) containing source/dest id, hop cnt and lifetime.

• Route replies return along reverse path – nodes update routing tables to destination. – After “lifetime” routes will be dropped from memory.

Page 13: Topic 3: Sensor Networks and RFIDs

AODV

Variations:– Route maintenance: neighbors send “hello” messages

and propagate route error messages when links break.– Other route metrics: replace hop cnt with some other

metrics.– Limited flooding: include max. hop cnt in RREQ.

Page 14: Topic 3: Sensor Networks and RFIDs

Geographic routing

• When nodes have location-information can use this to facilitate routing.– Information may be from GPS, in-network localization,

or given at deployment.

• Geographic routing algorithms route only using location-information (I.e. no node ID’s)

• Natural correspondence to location-based queries.

Page 15: Topic 3: Sensor Networks and RFIDs

Geographic routing

• Simplest approach – “greedy forwarding”– Always send packet to neighbor who is closest to the

destination.

– Several variants:• E.g. nearest neighbor with forward progress.

– Problem?

Page 16: Topic 3: Sensor Networks and RFIDs

Geographic routing

• Getting around “dead-ends:”– “Face routing:” use a right-hand rule when stuck.

• Greedy perimeter stateless routing (GPSR) is an algorithm that uses these ideas.– Greedy routing until “stuck”

– Then face routing until “un-stuck”

Page 17: Topic 3: Sensor Networks and RFIDs

Data centric routing

• In many sensor network applications do not want to send data to specific nodes, but rather to any node that knows “x”.

• Traditional approach to such problems.– Add “higher layer names” for data of interest.– Somehow bind those names to the nodes which have that

data (e.g. thru a directory service)

• This leads to several levels of indirection, each which add overhead to the sensor network.– Moreover, since most sensor networks are application

specific, it is not clear this is needed.

Page 18: Topic 3: Sensor Networks and RFIDs

Data-centric routing

• Main idea: implement network operations directly using content names.

• Requires having a system for content-based addressing.– E.g. [Temperature, location, time].

• Need mechanisms for advertising & requesting content – Publish/subscribe model– Push vs. pull.

Page 19: Topic 3: Sensor Networks and RFIDs

Directed Diffusion

• One broad data centric routing approach is directed diffusion.– Inspired by ants.

• “Pull” algorithm – sinks transmit interests for certain types of data.– Interests are flooded through network.

• Other nodes cache the interest, and create a gradient towards the source – direction plus value.

Page 20: Topic 3: Sensor Networks and RFIDs

Directed Diffusion

• When a node obtains data that matches an interest it sends it along gradient direction– Rate (probability) proportional to gradient

value.

• Sink reinforces particular route(s)– Increase rate of gradient– Reinforcements propagate back to source.

Page 21: Topic 3: Sensor Networks and RFIDs

Directed Diffusion

• Quite general algorithm.– Probabilities can depend on energy/delay.– Multi-path version possible.– Also “push”-based versions.

Page 22: Topic 3: Sensor Networks and RFIDs

Outline

• Discuss applications in HW.• Loose ends from last time• Networking issues:

– MAC protocols– Topology formation– Routing

• Security• Future trends

Page 23: Topic 3: Sensor Networks and RFIDs

Security Concerns

• Access control

• Message integrity

• Message confidentiality

• Replay protection

• Denial of service protection– Jamming, energy depletion, etc.

Page 24: Topic 3: Sensor Networks and RFIDs

Security Concerns

• Access control

• Message integrity

• Message confidentiality

• Replay protection

• Denial of service protection

Addressed as optionsin 802.15.4/Zigbee

Page 25: Topic 3: Sensor Networks and RFIDs

Zigbee Security solutions

• Access control – coordinator maintains access list.

• Message integrity.– Include a message authentication code with

each (MAC layer) packet.– Cryptographic checksum– 4/8/16 byte options.

Page 26: Topic 3: Sensor Networks and RFIDs

Zigbee Security solutions

• Message confidentiality.– Encrypt message.

• Replay protection.– Include a sequence number in each packet.

Page 27: Topic 3: Sensor Networks and RFIDs

Today

• Discuss applications in HW.• Loose ends from last time• Networking issues:

– MAC protocols– Topology formation– Routing

• Security• Future trends

Page 28: Topic 3: Sensor Networks and RFIDs

Future Trends

• Markets

• Challenges

Page 29: Topic 3: Sensor Networks and RFIDs

Markets

“By 2008, there could be 100 million wireless sensors in use, up from about 200,000 today. The worldwide market for wireless sensors, will grow from $100 million this year to more than $1 billion by 2009.”

– Harbor research, 2006

Page 30: Topic 3: Sensor Networks and RFIDs

Markets

Many companies are positioning themselves at several different points in the sensor network “value chain”: – Component suppliers (e.g. chips, mini-controllers, sensors)

• Freescale, Intel, TI (Chipcon), ….

– Network technology companies (e.g. motes, access points)• Crossbow, Millennial Nets, Dust Networks,…

– System integrators (installation services, applications)• Intel, IBM, HP,…

– Data management/analysis (what to do with all this data)• Oracle, SAP, Microsoft,…

– Also Companies with specialized Apps.• Home automation, toys, …

Page 31: Topic 3: Sensor Networks and RFIDs

Markets

• RFID is already seeing widespread deployment.• Some Zigbee products are also becoming available

for applications such as home automation.• Markets for other sensor net applications are also

emerging.• However, some of the more exotic apps are still

not feasible.– Large deployments of low power nodes, with multi-hop

routing and long-lifetimes.

Page 32: Topic 3: Sensor Networks and RFIDs

Challenges

• Reliability (lifetimes).– Made a lot of progress here, but improvements still

needed in some areas.

– As we have seen, some will come from Moore’s law - others will require new solutions.

• Security/Privacy– Societal implications of widespread use of

sensors/RFIDs

– Privacy/regulation/societal acceptance.