BlueTooth ScatterNet Performance Simulator BSPSOctober 2001

1
BlueTooth ScatterNet Performance Simulator

• Supervisor Gil Zussman
• By Liron Har-Shai Ronen Kofman

2
Agenda

• BlueTooth ScatterNet
• ScatterNet Structure in OPNET
• BlueTooth node Implementation
• Results
• For PicoNet model
• For ScatterNet model
• Plans for the future

3
BlueTooth Scatternet

• PicoNets are created ad-hoc
• PicoNet consists of one master and up to seven
  slaves and works in a Master-Slave concept.
• ScatterNet is created by joining two PicoNets
  with a bridge
• Bridge is a BT node that
• functions as slave in both
• PicoNets and jumps
• between them.

4
Routing in ScatterNet

• How will the bridge share its time between
  PicoNets ?
• How will a PicoNet Master treat the bridge ?
  (priority)
• What BlueTooth modes should the bridge use ?
  (park/hold/sniff)

5
ScatterNet Structure in OPNET
All the models were made from scratch in order to
fit the BlueTooth specifications
6
BlueTooth Node Implementation

• 1 radio transmitter (rt_0)
• 1 radio receiver (rr_0)
• 4 main Processes
• Poisson packet source
• Queue
• Application
• BaseBand
• 1 infinite buffer for packets

Packet Flow
7
BaseBand Process Implementation
check AM_ADDR
8
Application Process Implementation (MASTER)
Receive the packet from the lower layer and
extracting the relevant fields
Preparing the packet for sending and analyzing
results
9
Application Process Implementation (SLAVE)
10
Results - PicoNet

• In order to validate our model we measured the
  average delay in a PicoNet
• The PicoNet is using Strict Round Robin policy
• The Master polls each Slave in its turn every
  round

11
Strict Round Robin Timing Scheme

• We can see the Strict Round Robin Policy as
  implemented in our model (1 Master and 4 Slaves)

Master
12
Average Delay (Slave to Master)

• n Number of Slaves.
• ? - Arrival rate (packet/sec).
• M Number of slots per frame.
• T Packet transmission time.
• Tc Cycle time.
• ? Load Factor.
• Wq Expected queuing time of a packet.
• D Expected packet delay.
•  
• M 2n
• T 0.0625 Seconds
• Tc MT 2nT
• ? ? Tc

13
Average Delay (Slave to Master)

• n Number of Slaves.
• ? - Arrival rate (packet/sec).
• M Number of slots per frame.
• T Packet transmission time.
• Tc Cycle time.
• ? Load Factor.
• Wq Expected queuing time of a packet.
• D Expected packet delay.
•  
• M 2n
• T 0.0625 Seconds
• Tc MT 2nT
• ? ? Tc

14
Results PicoNet (cont.)

• We created a mechanism that enable us to measure
  average delay between slaves within the PicoNet.
• The PicoNet is using strict Round Robin policy
• We are measuring the average delay between Slave1
  and Slave2 to Slave3.
• All the packets are in the same size (1 slot
  time).

15
Average Delay (Slave to Slave)
Average delay from Slave1 to Slave3
Average delay from Slave2 to Slave3

• We can see that the Strict Round Robin Policy
  doesnt necessary cause similar delay between
  different sources.
• The delay from the Slave1 to Slave3 is bigger
  than the delay from Slave2 to Slave3 in 1 slot.

16
Results - ScatterNet
PicoNet 1
PicoNet 2
Bridge

• Our goal was to create an initial ScatterNet
  model which will be base for further improvements
• The bridge is jumping between the 2 PicoNets in a
  constant rate
• There is NO-RESPONSE Mechanism that enables the
  Master to track which slaves are connected to him
  and to avoid critical timing dependencies
• We are using a Prioritized Round Robin. If the
  bridge is in the other PicoNet its priority is
  decreased

17
BlueTooth Bridge Implementation

• 2 radio transmitter (rt_0)
• 2 radio receiver (rr_0)
• 5 main Processes
• Poisson packet source
• Queue
• Application
• BaseBand
• LMP
• 1 infinite buffer for packets that the bridge
  produce
• 2 infinite Queues one for each PicoNet that the
  bridge is connected to (2 PicoNets here)
  ROUTING

Packet Flow
18
Bridge Timing Diagram
Bridge is connected to PicoNet 1
Bridge is connected to PicoNet 2
We can see that the turn around time in PicoNet 1
is longer than PicoNet 2 because of the fewer
participants in PicoNet 2.
19
MASTER-BRIDGE Relationship in ScatterNet

• The NO RESPONSE Mechanism takes care of the
  scenario in which the Bridge is not connected to
  the Master.
• The Master re-enters the lost packet again to the
  head of the queue when NO RESPONSE occurs and in
  this way packets are not being lost.
• Average number of packets in queue over the
  period of time in which the Bridge is not
  connected stays the same.

20
Bridge Packet Switching
Bridge is connected to PicoNet 1
Bridge is connected to PicoNet 2
We can see that when the bridge is connected to
one PicoNet, the number of packets in the Queue
to this PicoNet is decreasing and when the bridge
is not connected, the number of packets in the
Queue is increasing.
21
Routing Mechanisms

• Every packet is stamped with the following
  fields
• Source BD_ADDR
• Destination BD_ADDR
• Creation time
• Other BlueTooth overhead information
• The main idea is to learn as much as possible
  about the ScatterNet topology.

22
Whats Next

• Analyzing packet delay over ScatterNet with
  various PicoNet policies
• Optimize bridge functionality to minimize delay
• Make the model as generic as possible