AdHoc Probe: Path Capacity AdHoc Probe: Path Capacity Probing in Wireless Ad Hoc NetworksProbing in Wireless Ad Hoc Networks
Ling-Jyh Chen, Tony Sun, Guang Yang, Ling-Jyh Chen, Tony Sun, Guang Yang,
M.Y. Sanadidi, Mario GerlaM.Y. Sanadidi, Mario Gerla
Computer Science Department, UCLAComputer Science Department, UCLA
Why Path Capacity?
Why do we want to measure path cap?
To adjust video rates; adapt end to end encoding
To select TCP parameters, etc
Example ScenarioExample Scenario Server is streaming video to user roaming in ad hoc net a shopping mall with “opportunistic” ad hoc extensions Assume 802.11g; path capacity varies from 2-54Mbps If user walks outside of 802.11 area, capacity drops to
GPRS (< 100Kbps) Server must know capacity to adjust video rate and avoid
network overflow!!
CapProbe CapProbe (Rohit et al, SIGCOMM’04)(Rohit et al, SIGCOMM’04) Key insight: a packet pair that gets through with zero
queueing delay on both packets yields the exact estimate Equivalently: zero queues -> Delay Sum Min -> exact CAP CapProbe uses “Minimum Delay Sum” filter
CapacityCapacity
Ad hoc path capacityAd hoc path capacity
The capacity of an ad hoc path is related to link speeds in a different way than on a wired path
This is because consecutive transmissions interfere with each other
Ad Hoc Path capacity definition the data rate achieved by a UDP stream on
the unloaded path (no other traffic) - this is the general definition
In the ad hoc path, path capacity = “narrow neighborhood” capacity (different from “narrow link” capacity in wired net)
Ad hoc path capacity (cont)Ad hoc path capacity (cont)
Ad Hoc Neighborhood The minimal set of nodes that must be inactive (no
tx nor receive) while a transmission takes place Equivalently, the region affected by the transmission Only one pkt can be in the neighborhood at a time
PP measures N-hood capacity N-hood capacity = (link speed)/(# of N-hood hops) The N-hood Capacity trivially reverts to link
capacity for the wired section of the path. Key result: The PP measurement yields the correct “path
capacity” regardless of wired, wireless or mixed configuration
Neighborhood Capacity
N-hood Cap in an ad hoc net can vary with: MAC protocol and link scheduling Link interference S/N ratio; Tx power Encoding/modulation scheme
Previous Work (Li et al, Mobicom 01)Previous Work (Li et al, Mobicom 01) Used UDP flow stream to probe the maximum
achievable throughput (brute force method) UDP file transfer may introduce excessive O/H
in some applications
1 hop1 hop 2 hop2 hop 3 hop3 hop 4 hop4 hop 5 hop5 hop 6 hop6 hop 7 hop7 hopAPAP
dispersion 2dispersion 2
sendersender
back to backback to back packets packets
wired Internet
wireless multihop
dispersion 1dispersion 1
Multihop path simulationMultihop path simulation
1500 byte pkts
Simulation of mobile hostsSimulation of mobile hosts Probing capacity of path
(n1 -> n6) n2~5 move clockwise 200 samples/run, 20 runs
Simulation of mobile end hostsSimulation of mobile end hosts Probing the capacity of path (0 ->25) Mobility: 1 m/sec; Cross Traffic: 1kbps/flow 200 samples/estimation; 4 samples/second
0
600 1200
18002200
2600 2800
3000
Testbed Measurements Testbed Measurements (WiTMeMo’05)(WiTMeMo’05)
Testbed configurations 802.11b fixed rate (2Mbps mode); chain
topology 802.11b auto rate; varying distance between
two nodes 802.11b auto rate; w/ Bluetooth interference 802.11b fixed rate (2Mbps mode); remote
probing from the Internet
Experiment Results (1)Experiment Results (1)
Fixed rate, variable hop length
Experiment Results (2)Experiment Results (2)
Auto Rate, variable distance
Experiment Results (4)Experiment Results (4) Probing from the Internet
SummarySummary
Wireless ad hoc Capacity estimation critical for Battlefield networks Emerging commercial ad hoc nets (eg
car2car) Ad Hoc Probe estimates correct e2e path
capacity of wired, wireless ad hoc and mixed nets .
User need not specify the environment or the MAC parameters
Simulation and measurements validate the findings
Next step: design the “adaptive” server