1. C. Lu, B.M. Blum, T.F. Abdelzaher, J.A. Stankovic, and T. He
1/1/2019
RAP:A Real-Time Communication Architecture for Large-Scale Wireless Sensor Networks
C. Lu, B.M. Blum, T.F. Abdelzaher, J.A. Stankovic, and T. He

2. Design Requirements Minimize end-to-end deadline miss ratio
Support distributed micro-sensing
High-level service API
Large scale, high density
Scalability is key
Extreme resource constraints
Minimal overheads
1/1/2019
Chenyang Lu

3. Location-based Communication
From location to location
What is the virus density in south terminal of airport?
Individual sensors NOT important
Local coordination: Sensors in interested area aggregate data
Sensor-base comm.: Send aggregated result to base station
ID-based
From ID to ID
What is the reading of sensor ?
Rely on (unreliable) individual sensors
1/1/2019
Chenyang Lu

4. RAP: Real-time locAtion-based Protocols
Sensing/Control
Application
Query/Event Service APIs
Query/Event Service
Coordination Service
Location-Addressed Protocol
Geographic Routing
Velocity Monotonic Scheduling
Prioritized MAC
1/1/2019
Chenyang Lu

5. Query/Event API RAP provides the following query/event service APIs.
Coordination Service
Location-Addressed Protocol
Geographic Forwarding
Velocity Monotonic Scheduling
Prioritized MAC
RAP provides the following query/event service APIs.
query { attribute_list, area, timing_constraints, querier_loc }
register_event { event, area, query }
Assume that the locations of the base stations are fixed.
1/1/2019
Chenyang Lu

6. Example .. register_event {
Query/Event Service
Example ..
Coordination Service
Location-Addressed Protocol
Geographic Forwarding
Velocity Monotonic Scheduling
Prioritized MAC
register_event {
virusFound(0,0,100,100), // area to post event
query { // query to be triggered
virus.count, // attribute
area=(x-1,y-1,x+1,y+1), // query area
period=1.5, deadline=5, // timing info
base=(100,100) // base station location }
The following API call registers a virus_count query for a virus_found event. If any viruses are found in a rectangular area with coordinates (0,0,100,100), returns the average density of the viruses of the 2×2 square area centered at the event location (Xevent,Yevent) every 1.5 sec. Every reading should reach the base station within an end-to-end deadline of 5 sec
1/1/2019
Chenyang Lu

7. query { virus.count, area=(10,10,12,12), period=1.5,deadline=5,
base=(100,100) }
The following query requires the average density of the viruses in an rectangular area (10,10,12,12) be reported to the base station of the querier every 1.5 sec. Every reading should reach the base station within an end-to-end deadline of 5 sec.
1/1/2019
Chenyang Lu

8. Location-Addressed Protocol
LAP is a connectionless transport layer in the network stack. LAP is similar to UDP except that all messages are addressed by location instead of IP address. Three types of communication are supported by LAP: unicast, area multicast, and area anycast.
Unicast delivers a message to a node that is closest to the destination location. Unicast can be used by sensors to send query results to base stations.
Area multicast delivers a message to every node in a specified area. Area multicast can be used to register for an event or send a query to an area, for coordination among nodes in a local group.
Area anycast delivers a message to at least one node in a specified area. Area anycast can also be used for sending a query to a node in an area. The node can initiate group formation and coordination in that area.
1/1/2019
Chenyang Lu

9. Geographic Forwarding
1/1/2019
Query/Event Service
Geographic Routing
Coordination Service
Location-Addressed Protocol
Geographic Forwarding
Velocity Monotonic Scheduling
Closest to C
Prioritized MAC A C E
What if there is holes/obstacles?
Local state  Scalability – Routing decisions are local
Dense network  Efficient greedy forwarding works well
Dense network  #hop proportional to distance
Location-based comm.  No location directory service
1/1/2019
Chenyang Lu

10. Background – GF
GF always chooses the node that is closest to the destination in FS. s d
1/1/2019
Chenyang Lu

11. Deadline & Distance Aware
Existing ad hoc networks, packets are typically forwarded in FCFS order. FCFS scheduling does not work well in real-time networks where packets have different end-to-end deadlines and distance constraints.
Deadline-aware means that a packet’s priority should relate to its deadline. The shorter the deadline, the higher the packet priority.
Distance-aware means that a packet’s priority should relate to its distance from the destination. The longer the distance, the higher the packet priority.
1/1/2019
Chenyang Lu

12. Velocity Monotonic Scheduling
Query/Event Service
Velocity
Coordination Service
Location-Addressed Protocol
Geographic Forwarding
Velocity Monotonic Scheduling
Prioritized MAC
Timing constraint: deadline
Location constraint: distance to destination
Requested Velocity
Embody both constraints
Reflect local urgency
Velocity Monotonic Scheduling (VMS):
Priority = Requested Velocity
1/1/2019
Chenyang Lu

13. Example D A C B dis = 90 m; D = 2 s V = 45 m/s HIGH Priority
LOW Priority
1/1/2019
Chenyang Lu

14. Velocity Monotonic Scheduling
Query/Event Service
Coordination Service
Velocity Monotonic Scheduling
Location-Addressed Protocol
Geographic Forwarding
Velocity Monotonic Scheduling
Prioritized MAC
Static VMS
Fixed velocity on each hop
V = dis(x0,y0,xd,yd)/D
Source location: (x0,y0)
Destination location: (xd,yd)
End-to-end deadline: D
Dynamic VMS
Adapt velocity at intermediate node based on progress
Vi = dis(xi,yi,xd,yd)/Si
Velocity at node: Vi
Location of node i: (xi,yi)
Slack: Si = D – elapseTime
1/1/2019
Chenyang Lu

15. Priority Queue Single Queue: Multiple Queue: Ordered by priority
Queue is full, higher priority incoming packets overwrite lower priority.
Implementing a data structure whose insertion time, in the worst case, grows logarithmically in the number of packets.
Multiple Queue:
Priority corresponds to a range of requested velocities. A packet is first mapped to a priority, and then inserted into the FIFO queue that corresponds to its priority
Packets that miss their deadlines are useless, priority queues actively drop packets that have missed their deadlines to avoid wasting bandwidth.
1/1/2019
Chenyang Lu

16. Velocity Monotonic Scheduling
1/1/2019
Query/Event Service
Prioritized MAC
Coordination Service
Location-Addressed Protocol
Geographic Forwarding
Velocity Monotonic Scheduling
Prioritized MAC
Collision Avoidance (CA)
Channel idle  wait for DIFS = BASE_DIFSPRI
Packets with a higher priority (corresponding to a smaller PRIORITY value) on average choose a smaller waiting period.
Contention
Collision (No CTS or No ACK)CW = CW*(2+(PRI-1)/MAXPRI)
MAXPRI is the maximum value of priority(corresponding to the lowest priority). The backoff counter of a node with a pending lower priority packet increases faster than a node with a pending packet with a higher priority.
Similar to ’s EDCF
Acquire
Channel
Idle
Time
BASE_DIFSPRI CW
Avoidance
Contention
Exponential Backoff
Transmission
1/1/2019
Chenyang Lu

17. Example … Count: registerEvent { virusFound(0,0,136,136), query {
virus.count,
area=(Xevent-1,Yevent-1,Xevent+1,Yevent+1),
period=Pc, deadline=Dc
base = (134.07, ) };
Detail: query {
detail,
area=(x-1,y-1,x+1,y+1),
period=Pd, deadline=Dd
base=(134.07, )
1/1/2019
Chenyang Lu

18. Example …
A user registers a count query with a virusFound event in the whole 136*136 m2 squared area. A virusFound event is generated when a grid detects a specified virus at location (Xevent,Yevent). This event triggers a query virus_count, which periodically reports the density of the detected virus in the area (Xevent-1,Yevent-1,Xevent+1,Yevent+1) to the base station
A user can also directly submit a detail query to get more detailed data collected at a location. Detail generates periodic flows that send detailed information about a grid to the base station for further analysis. While a
large number of count flows may be generated (e.g., during a bio-attack), the user may only query the details of a small number of important grids. We assume that packets (called count packets) returned by count queries have longer deadlines than packets (called detail packets) returned by detail queries.
1/1/2019
Chenyang Lu

19. Simulation: Biometric Sensing
1/1/2019
Simulation: Biometric Sensing
100 nodes on 136X136 m2
Periodic query count on 31 nodes; detail on 15 nodes
Base
Station
Hot Regions
(sources)
6-7 hop maximum
FAR
1/1/2019
Chenyang Lu

20. Workload Communication range: 30.5 m
Limitations of the GloMoSim simulator, we had to send data flows on top of the UDP/IP stack that contribute to 28 B overhead. In a real implementation we expect to eliminate the UDP/IP headers
Network (roughly approximate MICA mote)
Communication range: 30.5 m
Packet size: 32B (count), 160 B (detail)
Bandwidth: 200 kbps (> MICA)
Protocols
Routing: DSR (Dynamic Source Routing), GF (Geographic Forwarding)
Scheduling: FIFO, DS (Deadline-based), SVM, DVM
MAC: , extended w. prioritization
1/1/2019
Chenyang Lu

21. Details …
The main performance metric is the deadline miss ratio, i.e., the percentage of generated packets that are received by the base station within their deadlines
DS assigns a fixed priority to packets based on their end-to-end deadlines. In our workload, all count packets are assigned priority 3 (the lowest), and all detail packets are assigned priority 1 (the highest). It doesn’t depend on the distance.
DS and SVM perform almost identically when deadline is (5, 50) s. This conforms to our intuition. While the distances from each sensor to the base station stays the same, the bigger differences between the deadlines of detail and count flows become a dominant factor in requested velocity.
1/1/2019
Chenyang Lu

22. Flow of Packets Base station Base station GF – Flow of Packets
DSR – Flow of Packets
1/1/2019
Chenyang Lu

23. Deadline Miss Ratio Overall
1/1/2019
Deadline Miss Ratio Overall
GlomoSim simulation (deadline: detail: 5 s, count: 10 s)
1/1/2019
Chenyang Lu

24. Deadline Miss Ratio: FAR hot region
1/1/2019
Deadline Miss Ratio: FAR hot region
GlomoSim simulation (deadline: detail: 5 s, count: 10 s)
1/1/2019
Chenyang Lu

25. Distance Fairness SVM provides “fairer” service to remote sensors
Critical for scalability of sensor networks!
1/1/2019
Chenyang Lu

26. Conclusion Velocity Monotonic Scheduling
Reduce end-to-end deadline miss ratio
Fair service to remote sensors
Event/query service API’s
High-level abstraction for distributed microsensing
Location-based protocol stack
Scalable
Small protocol overhead
1/1/2019
Chenyang Lu

27. Critique
VMS
DVM
No schedulability analysis or admission control  No guarantee
Is velocity the right trade-off between distance and time?
Distance = #hop? in-doors, obstacles, void, long links
No routing or congestion control to handle congestion: RAP+SPEED+?
Location of the base station is fixed.
Lack support for
Coordination services
Area multicast
Mobility
1/1/2019
Chenyang Lu

28. Related Work
Directed diffusion is a data-driven communication paradigm for sensor networks. Users can broadcast interests to sensor networks. Sensors whose data match an interest report their data to the node that posts the interest
At the MAC layer Woo and Culler proposed a MAC protocol with adaptive rate control to achieve fairness among nodes regardless of their distance from the base station in a sensor network. However, their MAC does not provide prioritization for packets with different velocities. Timing constraints are not considered in their protocol.
1/1/2019
Chenyang Lu

29. Related Work …
VMS is inspired by coordinated multihop scheduling developed by Kanodia et. al. They proposed three priority index assignment policies for multi-hop wireless networks. The Time-To-Live (TTL) policy assigns priority to a packet based on its TTL counter, while each node decreases TTL by the time it spent in that node. The TTL-based priority can dynamically adapt packet priorities based on its progress. We note that TTL-based priority may not handle scenario 2 in Figure 2 well because A and B’s packets may have a similar TTL despite the fact that they have different distances to travel after E. The fixed per-node allocation decreases the priority index on each node by a per-node constant.
1/1/2019
Chenyang Lu

30. Related Work …
The uniform delay budget (UDB) allocation assigns a fixed priority index to a packet based on its end-to-end deadline divided by the end-to-end hop count. UDB essentially utilizes per-hop velocity computed based on end-to-end hop count, while VMS is based on geographic velocity computed based on the geographic distance to the destination. UDB requires routing protocols to provide the end-to-end hop count for each flow at the cost of route discovery and maintenance overhead. UDB cannot work with GF, which does not provide hop count. In comparison, VMS does not require hop count and is a perfect match with GF.
1/1/2019
Chenyang Lu