RAP and SPEED

1
RAP and SPEED

• Presented by Octav Chipara

2
Realtime Systems

• Concerned with two aspects
• Control
• RAP
• Predictability
• SPEED

3
Modeling the sensor networks

• A sensor
• Limited memory
• MAC layer may provide QoS
• Communication
• Scarce bandwidth
• Voids exists
• Energy intensive
• Communication generates congestion hot-spots
• end-to-end communication time depends on
  single-hop delay and the distance it has to travel

4
Modeling the sensor networks

• Tight integration with the physical world
• Location aware
• Communication patterns
• Local coordination
• Sensors coordinate with one another usually by
  defining a group in order to aggregate data
• Usually involves a small number of hops
• Sensor-base communication
• Sends data from the local group to the base
  station
• Requires multiple hops

5
RAP
6
Contributions

• A high level architecture for large wireless
  networks
• Ability to define queries
• Use event services
• Velocity Monotonic Scheduling (VMS)
• A policy for scheduling packets in a sensor
  network

7
Design Goals

• Provide APIs for micro-sensing and control
• Maximize the number of packets meeting their e2e
  deadlines
• Scale well to large number of nodes and hops
• Introduce minimum communication overhead

8
RAP Stack
9
Query/Event Service API

• Query attribute list, area, time constraints,
  querier location
• Register event, area, query
• Periodically the results will be send back to the
  querier

10
Location-Addressed Protocol

• Connectionless transport layer
• Address is based on geographic location
• Services
• Unicast
• Area multicast
• Area anycast

11
Geographic forwarding

• Greedy algorithm
• Selects the node with the shortest geographic
  distance to the packets destination
• At every step the packet gets closer to the
  destination
• Works really good for high density network

12
Velocity Monotonic Scheduling

• FCFS policy is generally used in sensor networks
• Works poorly for realtime systems
• VMS
• It is both deadline and distance aware
• Assigns priority based on the requested velocity
• A higher velocity denotes higher priority

13
Velocity Monotonic Scheduling(2)

• Two flavors
• Static monotonic velocity (SMV)
• Vdist(x0,y0,xd,yd)/D
• Dynamic monotonic velocity (DMV)
• Vdist(xi,yi,xd,yd)/(D-Tj)

14
MAC layer prioritization

• When communicating multiple host compete for the
  shared medium
• In 802.11b all messages have the same priority
• To enforce packet prioritization MAC protocols
  should provide distributed prioritization on
  packets from different nodes
• We can changed two parameters
• The time you can wait after idle
• DIFS BASE_DIFS PRIORITY
• Backoff increase function
• CWCW(2 (PRIORITY-1)/MAX_PRIORITY)

15
Experiments

• Routing protocols
• Dynamic Source Routing (DSR)
• Originally designed for ID networks
• GF
• Location based routing
• Scheduling
• FCFS
• DS fixed priority based on their e2e deadline
• Velocity Scheduling
• DVS
• SVS

16
Overall Miss Ratio of GF/DSR
17
Impact of scheduling on deadline misses
18
SPEED
19
State of the art

• Only a few real-time algorithms exist for sensor
  networks
• Routing based on sensors position
• GPSR
• GEDIR
• LAR

20
Design Goals

• State
• Information regarding the immediate neighbors
• Soft Real Time
• Provides uniform speed delivery across the
  network
• Q is this uniform speed guaranteed globally or
  piece-wise
• Q why would the application want something like
  this
• Minimum MAC Layer support

21
Design Goals

• Traffic load-balancing
• Localized behavior
• Void avoidance

22
Speed Protocol

• API
• Neighbor beacon exchange scheme
• Delay estimation algorithm
• Stateless nondeterministic geographic forwarding
  algorithm
• Neighbor feedback loop
• Backbone rerouting
• Void avoidance
• Last mile processing

23
API

• AreaMulticastSend(position,radius, packet)
• Send a message to all nodes in the specified area
• AreaAnyCastSend(position,radius,packet)
• Sends a message to at least one node inside the
  specified area
• UnicastSend(Global_ID, packet)
• Send a message to the specified ID (id is given
  by its geographic position)
• SpeedReceive()

24
API (2)

• Where is the time constraint?
• SPEED aims at providing a uniform packet
  delivery speed across the sensor network, so that
  the end-to-end delay of a packet is proportional
  to the distance between the source and
  destination. With this service, real-time
  applications can estimate end-to-end delay before
  making admission decisions. Delay differentiation
  for different classes of packets is left as
  future work.
• Q if the velocity is imposed piece wise how can
  we give end-to end guarantees?
• Q By distance we mean Euclidian distance or the
  Euclidian distance of the path?

25
Neighbor beacon exchange scheme

• Periodically broadcasts a beacon to neighbors to
  exchange local information
• In order to reduce traffic we can piggyback the
  information
• Possible enhancement
• Advertising state changes may reduce number of
  beacons
• On-demand beacons
• Delay estimation
• Back pressure pressure
• Fields
• Neighbor ID
• Position
• Send To Delay

26
Delay estimation algorithm

• Due to scarce bandwidth cannot use probe packets
• Delay is measured at the sender as the difference
  between when the packet was queued and its ACK
  and the processing time on the receiver time
• Keeps track of multiple data points to compute
  the current delay using (EWMA)
• intervalavg intervalavg w err
• Where
• I is the refresh time sample from this packet
• err I - intervalavg
• w is the smoothing constant

27
Delay estimation algorithm (2)

• Delay estimation beacon is used to communicate to
  other neighbors the estimated delay
• Hope to make a set of neighboring nodes react to
  changes in traffic patters avoid congestion

28
SNGF

• Neighbor set of node i
• Forwarding candidate set
• Where
• L d(i, destination)
• Lnext d(next, destination)
• Relay speed

29
SNFG(2)

• If (FSi gt 0)
• if (Viable gt 0)
• candidatechoose(Viable(FSi))
• send to candidate
• else
• compute relay ratio
• if no nodes to support Ssetpoint downstream,
  drop packet if a random chosen between (0,1) is
  bigger than the relay ratio
• else
• drop packets
• send pressure beacon upstream

30
SNFG(3)

• Delay Bound Le2e / Ssetpoint
• Where
• Le2e is the end-to-end Euclidian distance
  measured
• Ssetpoint the speed maintained across the network
• Drawbacks
• All messages have the same speed
• Does this really mean we can enforce a deadline?

31
Neighbor feedback loop

• Goal
• Maintain a single hop speed above a desired
  Ssetpoint
• Ssetpoint is a network wide parameter that tunes
  how harsh the realtime requirements are

32
Neighbor feedback loop
33
Back pressure rerouting

• Rerouting due to pressure
• The congested area is detected and the
  probability of sending to that node is limited
• Issue
• Maybe reinforcement should refer to a geographic
  area rather than a node!

34
Void avoidance

• Voids occur if the density is not high enough
• Deals with voids similarly to hotspots by
  applying backbone pressure
• Several packets may be dropped when trying to
  avoid a void

35
Last mile processing

• Processing close to the destination area
• Area anycast
• Area multicast

36
E2E Miss Ratio

• Ssetpoint 1km/s
• e2e deadline 200 ms

37
Protocol Evaluation
38
Improving Speed
39
Possible Improvements

• Allow the application to specify different speeds
• Combine speed and rap
• Change the MAC layer to support priority sending
  messages with different priorities
• What scheduling policy to use?
• Why are messages dropped?

40
Questions?