[attachment=15237]
Abstract
Most local area wireless technologies like IEEE 802.11 do not impose any restrictions on the topology of nodes and connections on which they will operate, but Bluetooth imposes some con¬straints for constructing valid topologies. The performance of Bluetooth over networks of a large number of nodes depends heavily on this topology, and in this paper we have evalu¬ated the criteria for making optimal topologies which maximize the total throughput that can be achieved in a Bluetooth network. We then pro-pose the Dynamic Scatternet Construction Pro¬tocol (DSCP) which constructs topologies on the fly, and specify routing schemes to work over the topologies constructed by DSCP. As a supplementary work, we have also com¬pared 802.11 and Bluetooth in the small distance Personal Area Networking scenario, and we find that Bluetooth holds a promising future in this arena since it has a greater degree of scalabil-ity in crowded networks of small area. We have also proposed a synchronization protocol which can bring about a desired synchronized pattern along multi-hop routes of data transfers. This synchronization will be important in the efficient utilization of the offered capacity.
Keywords: Bluetooth, Piconets, Scatternets, Bluescat, NS-2, DSCP
1 Introduction
Bluetooth [1] is a short range wireless tech¬nology intended to replace the cable(s) connect¬ing portable and fixed devices. It operates in the unlicensed ISM band at 2.4 GHz, and uses a frequency hopping TDD (Time Division Du¬plex) scheme for transmission. The maximum bandwidth possible is 1 Mbps. On the channel, each packet is transmitted on a different hop fre¬quency.
The Bluetooth system provides a point-to-point connection (using two Bluetooth devices), or a point-to-multipoint connection (using a maxi¬mum of eight Bluetooth devices). Two or more devices sharing the same channel form a piconet, with one device acting as the master of the pi¬conet, and the other device(s) acting as slaves. At the most seven active slaves can remain at¬tached to a master at any instant. In all cases, the master controls the channel access. Multiple piconets with overlapping coverage ar¬eas form a scatternet. Each piconet can only have a single master. However, slaves can par¬ticipate in different piconets on a time-division multiplex basis. In addition, a master in one piconet can be a slave in another piconet. Dif¬ferent piconets are not time or frequency syn¬chronized, and each piconet has its own hopping channel. The hopping sequence of frequencies in a piconet is a function of MAC address of the master. Therefore, no two piconets can have the same hopping sequence. The presence of scatternets becomes imperative when the num¬ber of active Bluetooth devices exceeds eight, and some slaves/masters have to act as bridging units to page link the different piconets together. The bridge nodes are in HOLD mode in one piconet and CONNECTION mode in the other piconet. They switch between their two nodes in both pi¬conets at the same time.
There are multiple ways of scatternet topology construction on the same ad hoc collection of nodes [6], and the overall network performance is greatly effected by the scatternet topology. For example, dispersed topologies with fewer number of bridges can lead to bottleneck bridges. Sim¬ilarly, dense topologies with a large number of bridges can lead to wastage of bandwidth ca¬pacity of the bridges. Therefore, topology con¬traction becomes an orthogonal research issue. BTCP [5] was proposed as a scatternet topol¬ogy construction protocol which connects nodes starting at the same time in a strongly connected network. However, in real life scenarios, nodes will hardly ever be switched ON at the same time. Therefore a dynamic topology construc¬tion scheme is needed which makes provision for realtime addition and deletion of nodes. In this paper we have proposed DSCP (Dynamic Scat¬ternet Construction Protocol) which maximizes the overall network capacity in dynamic scenar¬ios.
In section 2 we have described the simulation tool which we have used [4]. We then describe our experiments and results on optimal topolo¬gies maximizing the total network throughput in
Section 3, followed by a description of the DSCP protocol in Section 4. Finally we present our conclusions and describe the work that we are currently pursuing as an extension to our exper¬iments. We describe some additional work that we have done on improving scatternet capacity in Appendix-A and Appendix-B.