Internet Scheduling of Bluetooth Scatter Nets By Inter Piconet Scheduling

Integrated Intelligent Research (IIR) International Journal of Web Technology
Volume: 02 Issue: 01 June 2013 Page No.6-8
ISSN: 2278-2389
6
Internet Scheduling of Bluetooth Scatter Nets By
Inter Piconet Scheduling
S P Vijayaragavan1
,Chandan kumar2
, K.Ashok Augustine Raj3
1
Asst professor, Dept of EEE , Bharath University, Chennai, India
2
Dept of EEE , Bharath University, Chennai, India
3
Asst professor ,Dept of computer science , Bharath University, Chennai, India
Abstract-Bluetooth includes the concept of devices
participating in multiple ”piconets” interconnected via bridge
devices and thereby forming a ”scatternet”. This paper presents
a scheme for Bluetooth scatternet operation that adapts to
varying traffic patterns. According to the traffic information of
all masters that the bridge is connected, the bridge switches to
the masters with high traffic loads and increase the usage of the
bridge. In addition, load adaptive interpiconet scheduling can
reduce the number of failed unsniffs and the overhead of the
bridge switch wastes to further increase the overall system
performance.
I. INTRODUCTION
Bluetooth is a technology used for wireless personal area
networks. Bluetooth is low cost, radio-based wireless network
technology providing a limited coverage area without the need
for infrastructure. A piconet is a basic structure in Bluetooth,
where a piconet is a collection of Bluetooth devices, which are
synchronized to the same hopping sequence[7]. Each piconet
has exactly one master and up to 7 active slaves, where the
master administers the whole piconet[2]. One device in the
piconet can act as the master, all other devices connected to the
master must act as slaves. The unit establishing the piconet
automatically becomes the master, all other devices will be
slaves. The master determines the hopping pattern in the
piconet and the slaves has to synchronize to this pattern. Each
piconet has a unique hopping pattern. If a device wants to
participate it has to synchronize to this hopping pattern. A
slave may connect to more than one master. The slave
connecting to two or more masters is termed as a bridge. A set
of piconets that are interconnected by bridges is referred as a
scatternet. Although a bridge can participate in two or more
piconets, it can only serve in one piconet at a time. The bridge
will switch among all piconets it is connected in a time-sharing
fashion. The master that the bridge is serving is called the
serving master and the other masters that the bridge is
connected but not in service is called waiting masters.
This paper is organized as follows. Section 2 gives the
overview about load adaptive switching and Section 3
describes the model for load adaptive interpiconet scheduling.
Section 4 about the phases involved while the bridge switch
problem and its solution is proposed in Section 5. Finally
Section 6 concludes the paper.
II. OVERVIEW OF LOAD ADAPTIVE
INTERPICONET SCHEDULING
Load adaptive interpiconet scheduling of Bluetooth
Scatternet is presented in this paper. In this, a bridge will
switch to another piconet only when the current serving master
notices it to switch off[3]. As a result, there will not be any
packet loss when the bridge switches to another piconet[1].
Each master will maintain a scheduling table. The table records
all masters’ traffic information and their bridge usage statuses,
such as how long the master has waited for the usage of the
bridge, how long the master may not use the bridge, and so on.
The scheduling table of the serving master will be transferred
to the new serving master through the bridge when the serving
master decides to release the usage of the bridge. Based on the
scheduling table, the serving master can predict the time it may
not get the usage of the bridge after it releases the usage of the
bridge. After releasing the bridge, the master will not unsniff
the bridge during the time interval it has predicted. Therefore,
the number of failed unsniffs can be much reduced. When the
new serving master gets the scheduling table from the bridge, it
can figure out the minimum time it can freely use the bridge.
The bridge can dynamically switch among the piconets that it
is connected according to the master’s traffic loads and waiting
time.The power saving mode that a bridge uses to switch
among piconets directly influences the performance of the
scatternet[6]. Bluetooth specifies three power saving modes,
sniff, hold, and park modes[4]. In general, sniff mode is used
for a bridge to switch among piconets regularly and
periodically[1]. A device in sniff mode only wakes up
periodically in pre-arranged sniff slots. The master and the
slave must negotiate the sniff timing information, such as the
first sniff slot, sniff interval TSniff , sniff attempt, and the sniff
timeout. The sniffing slave only listens for the traffic during
the sniff slots. If no message is addressed to the sniffing slave,
the sniffing slave ceases listening for packets. If a message is
received in a sniff slot, the sniffing slave continues listening
for further sniff timeout slots after the sniff slot. In other
words, the transmission time is flexible between the master and
the sniffing slave. The master and the sniffing slave can only
communicate with each other in their pre-scheduled sniff slots.
If either one can not receive packets from the other in a sniff
slot, it is called a failed unsniff. A failed unsniff will lead to
one packet loss. Hence, too many failed unsniff will
significantly degrade the performance of the piconet, even that
of a scatternet.Challenges for Load Adaptive Switching are
• The variance of traffic loads. Since the traffic of a network
must be variable, therefore, load adaptive switching must have
the flexibility to dynamically adjust to meet the variant traffic
loads.
• The tradeoff between the throughput and the
transmission delay. To achieve the maximum throughput of a
scatternet, we would like to allocate more bridge service time
to the master with high traffic loads. However, it may increase
the transmission delay of the masters with low traffic loads.
Therefore, in addition to increase the throughput, to reduce
transmission delay is also needed to be considered.

ISSN: 2278-2389
7
• The frequency of a bridge switching among piconets.
When a bridge switches to a new piconet, the bridge may not
match to the timing of the new piconet. The bridge has to wait
until the next even slot to be unsniffed by the new serving
master. The time that a bridge waits for being unsniffed after
switching to a new piconet is called guard time waste.
Reducing the switching frequency of a bridge among piconets
can reduce the guard time waste[5].
III. MODEL FOR LOAD ADAPTIVE INERPICONET
SCHEDULING
Load adaptive interpiconet scheduling is operated on a
constructed scatternet. The connection between a master and a
slave is ACL link only. In this, the sniff mode is used as the
operating mode for the bridge to switch among piconets. The
sniff interval negotiated by a bridge with its serving master is
TSniff . Here, each master will maintain a scheduling table,
which indicates the traffic information of all masters that
bridge is connected. According to the scheduling table, the new
serving master can figure out the time it can use the bridge and
the waiting master can calculate the time it needs not to poll
the bridge in the following sniff slots. A scheduling table with
field MID represents the identity of the master and the other
fields are described as follows.Queue Consuming Time(QT): is
defined as the time that a link still needs the bridge to serve. It
means the time that all the data packets in the queues of the
master and the bridge need to be transmitted completely. There
is a queue agent to monitor the status of the queue on either
side of a link. The bridge will notify the master about this
information at each communication with the master. Therefore,
the master can compute the QT. Lost Time(LT): is defined as
the time that a master can not get the usage of the bridge. Since
the scheduling table has the QTs of all masters that the bridge
is connected, when the serving master has to release the usage
of the bridge, according to QTs, it could predict the time that it
may lose the bridge after it releases the usage of the bridge.
The time is called LT. LT could be used to reduce the number
of failed unsniffs of the waiting masters. Waiting Time(WT): is
the time that the master has been waiting for the usage of the
bridge since it released the usage of the bridge. The historical
information q represents the history of the traffic loads. Since
the decision of the master releasing the bridge depends much
on the value of QT, the precision of QT will influence the
performance of load adaptive switching. Therefore, to obtain
QT precisely, the history of traffic loads is counted to evaluate
QT due to the temporal locality of the traffic. Let p be the
increment of the traffic in queue during a fixed time period, say
T. The queue agent responds to maintain p. Thus, q can be
obtained as q = p/T , q will be computed for every T time
period. After q is obtained, the queue agent will reset p to zero.
When the serving master has to release the usage of the bridge,
it would record the q in the scheduling table. Hence, when the
new serving master gets the usage of the bridge, it could
evaluate the QT more precisely for some waiting master.When
the serving master i has to release the usage of the bridge, it
will find a candidate to be the new serving master, say j, and
update the LTi. The serving master i first finds the minimum
LTj from the scheduling table, for some j. If there are more
than one minimum LT, then it selects the one with the
maximum WT. It means that the waiting master j has the
highest priority to get the usage of the bridge once the serving
master releases the bridge. The serving master has to update
LTi once it decides to release the usage of the bridge to the new
serving master j. However, QTj in the scheduling table of
master i is an out-of-date value since it is recorded when the
master j has released the usage of the bridge. Therefore, it does
not stand for the current traffic loads of the master j. As a
result, we can use qj to estimate the QTj roughly. Therefore, at
least, the time that the serving master i will not get the usage of
the bridge can be obtained as below, say WS.
WS = QTj + qj *WTj
On the other hand, to avoid the excessive transmission delay of
the waiting masters, a waiting threshold (Wthold) is used to
limit the transmission delay. If the WS is bigger than Wthold,
then WS is set to Wthold. Due to a master can communicate
with a bridge only on the sniff slots, WS may not coincide with
the sniff slots. So we have to add an offset, b, to match the
sniff slot exactly. b can be calculated as follows.
b = TSniff - ((WS + 2) mod TSniff ).
Therefore, the time that the serving master i will not get the
bridge after it releases the usage of the bridge is
LTi = WS + b
IV. PHASES INVOLVED IN LOAD ADAPTIVE
INTERPICONET SCHEDULING
This consists of two phases: bridge phase and bridgeless
phase. The serving master executes bridge phase and all the
other waiting masters perform bridgeless phase.
Bridge Phase
If a serving master i gets the usage of the bridge, it first finds
the minimum LTj from the scheduling table, for some j.
According to this information, master i can realize how much
time it can use the bridge freely. Besides, master i should be
responsible for the maintenance of the scheduling table. That
is, the serving master i should add 1 to each WT and subtract 1
from each LT in the scheduling table per slot. When LTj = 0,
master i must check if it has to release the usage of the bridge
to the waiting master j. When the released condition is
satisfied, the serving master i has to release the usage of the
bridge to the waiting master j. The serving master i then
performs the serving master part of the Bridge Release
Procedure. As described above, once the serving master i
intends to release the usage of the bridge, the serving master i
will calculate LTi by means of the scheduling table. After the
LTi is calculated, master i updates the LTi into the scheduling
table and resets the WTi to zero. Master i then transmits the
scheduling table to the bridge and informs the bridge to serve
the new serving master j. The role of the master i is turned
from the serving master to the waiting master. Thus, master i
then performs the bridgeless phase afterward. The bridge
receiving the scheduling table will perform the Bridge Release
Procedure as well, but the bridge part. The bridge then waits
for being unsniffed by the new serving master j and maintains
the scheduling table during this waiting period. The
maintenance is that the bridge would record the time slot count
(sc) since it got the scheduling table to receive the unsniff
message from the new serving master. When the bridge is
unsniffed by the new serving master, it would subtract sc from
each LT, add sc to each WT in the scheduling table, and then
transmit the scheduling table to the new serving master.

ISSN: 2278-2389
8
There are two conditions that the serving master i has to
release the usage of the bridge to the waiting master j.
1. WTj >Wthold, (TIME event),
2. QTj+qj *WTj > QTi+QCthold, (QUEUE event).
The first condition implies that the master j has been waiting
for the usage of the bridge over the Wthold. The second
condition implies that all the data needed to be transmitted
completely between master j and the bridge is bigger than
those between master i and the bridge plus a QCthold. The
Qcthold is designed for avoiding the ping-pong effect when
QTi and QTj are close to each other. The first condition is to
avoid the excessive transmission delay of the waiting master.
The second condition is to allocate more service time to the
link with high traffic loads. If none of the two conditions is
satisfied, then the serving master i could keep using the bridge.
To improve the throughput of a scatternet, the master with high
traffic loads will be allocated more service time.
Bridgeless Phase
The waiting masters that do not get the usage of the bridge will
perform the bridgeless phase. For some waiting master, say j,
according to the LTj that was calculated when the master j had
released the usage of the bridge, the master j could realize the
time it might not get the usage of the bridge. Therefore, the
waiting master j won’t unsniff the bridge during LTj . Hence, it
could reduce the number of failed unsniffs.
V. BRIDGE SWITCH PROBLEM
If a bridge switches to a piconet, but the serving master has no
data to the bridge. A Poll-Null sequence event happens. It is a
bridge switch waste. However, if the Poll-Null sequence event
is happened in the sniff slot between the master and the bridge,
it implies that this switch of the bridge is waste once. It would
lead the bridge to go back to sleep and the other waiting
masters would be unable to unsniff the bridge successfully.
This will reduce the usage of the bridge. In order to reduce the
number of bridge switch wastes, LT will be increased as long
as the bridge switch waste happens. When a bridge switches to
a serving master, the bridge will transfer the scheduling table
to the serving master on the first odd slot. If the serving master
receives from the bridge a DH1 ACK packet that no data is
included, except the scheduling table, the serving master will
regard the ACK packet as a Null packet. Thus, LT of the
serving master will be increased accordingly. An additional
time (PNT, Poll-Null time) is added to the LTi. As a result, the
LTi will be lengthened after the Poll-Null sequence event.
Thus, it could reduce the number of the bridge switch wastes.
A Poll-Null event counter is used to record the times of
successive bridge switch wastes, and the PNT can be obtained
as follows.
PNT = TSniff * 2( Poll-Null event counter )
The LTi will be lengthened after a bridge switch waste as
below.
LTi = LTi + PNT.
LT is getting larger while the bridge switch wastes happen
often. To avoid excessive transmission delay while master i has
data to the bridge, an upper bound, MaxLT, is to limit the
increasing of LT.
VI. CONCLUSION
In this paper, load adaptive interpiconet scheduling
dynamically adjust the bridge service time according to the
master’s traffic loads, reduce the number of failed unsniffs, and
further increase the system throughput. The primary idea is to
allocate the bridge service time to the master which needs the
most. That is, load adaptive switch would allocate enough
bridge service time to the master with high traffic loads and
reduce the bridge switch wastes. On the other hand, to avoid
excessive transmission delay of the master with low traffic
loads, this would allocate the bridge service time to a master
once the master has waited for a period of time, no longer than
Wthold. Moreover, the masters in bridgeless phase will reduce
the number of failed unsniffs because of the LT. To improve
the bridge switch problem, this will lengthen the LT after a
bridge switch waste and hence can reduce the number of the
bridge switch wastes. Comprehensively, this provides an
optimal performance especially under the environment with
various traffic loads.
REFERENCES
[1] S. Baatz, M. Frank, C. K¨uhl, P. Martini, and C. Scholz, “Adaptive
Scatternet Support for Bluetooth using Sniff Mode,” Proceedings of the
IEEE
[2] Conference on Local Computer Networks, November 2001.
[3] M. Kalia, S. Garg, and R. Shorey, “Scatternet Structure and Inter-
Piconet Communication in the Bluetooth System,” IEEE National
Conference onCommunications, New Delhi, 2000.
[4] G. Miklos, A. Racz, Z. Turanyi, A. Valko, and P. Johansson,
“Performance Aspects of Bluetooth Scatternet Formation,” First Annual
Workshop on Mobile and Ad Hoc Networking and Computing
(MobiHOC),2000.
[5] A. Das, A. Ghose, A. Razdan, H. Saran, and R. Shorey, “Enhancing
Performance of Asynchronous Data Traffic over the Bluetooth Wireless
Ad-hoc Network,” in Proceedings of IEEE INFOCOM 2001, April
2001,Cables. Prentice Hall PTR, 2001.
[6] W. Zhang and G. Cao, “A flexible scatternet-wide scheduling algorithm
for bluetooth networks,” in Proceedings of the 21st IEEE International
Performance, Computing, and Communications Conference, 2002.
[7] T. Melodia and F. Cuomo, “Ad hoc networking with Bluetooth: Key
metrics and distributed protocols for scatternet formation,” Ad
HocNetworks, Apr. 2001.
[8] R. Kapoor, M. Y. M. Sanadidi, and M. Gerla, “An analysis of Bluetooth
scatternet topologies,” in Proc. IEEE Int. Conf. Communications , 2002.

Internet Scheduling of Bluetooth Scatter Nets By Inter Piconet Scheduling

More Related Content

Similar to Internet Scheduling of Bluetooth Scatter Nets By Inter Piconet Scheduling (20)

More from ijbuiiir1 (20)

Recently uploaded (20)

Internet Scheduling of Bluetooth Scatter Nets By Inter Piconet Scheduling