1. Bluetooth Scatternets: An Enhanced Adaptive Scheduling Schema Simon Baatz Matthias Frank Carmen Kühl Peter Martini Christoph Scholz presented by Metin Tekkalmaz

2. Outline Introduction Related Bluetooth Topics Challenges in Scatternet Formation Credit Based Approach Performance Enhancements Conclusion

3. Introduction The Goal is to schedule the data traffic in the scatternet Paper discusses one of the ways of this which, Adapts to varying traffic patterns Is based on sniff mode Does not need much modifications on current spec

4. Related Bluetooth Topics Poll - Null messages Maximum Poll interval Sniff Mode

5. Challanges (1/3) Scatternets need some nodes to be connected with multiple piconets different hop sequences different timing Single transceiver: Single hopping schema at a time Switching is a necessity - in TDM

6. Challanges (2/3) Different slot boundaries High vs. Low switching frequency Throughput vs. Delay

7. Challanges (3/3) Scheduling Switches Switching but, when and to where How the coordination is achieved? Is “Presence Schedule” a solution?

8. Credit Based Approach (1/5) Presence Point Concept (let’s call it PP) “Presence Points” (o/l) instead of “Presence Schedules” (a priori) PP is where communication on a link (between Master & Slave) may start Quickly determine if peer is available Begin communication if available Try another PP otherwise Communication period is not predetermined

9. Credit Based Approach (2/5) PP implementation w/ Sniff Mode Let the sniff slots be the PPs Maximize probability of a common sniff slot Continue sniff event until one or both sides decide to abort Max. poll interval provides QoS When and Where problem is still unsolved

10. Credit Based Approach (3/5) The Solution: Credit Schema Devices assign priorities to each link, locally Priority is per peer, not per link i.e. Different devices may assign different priorities to the same link A link has a higher priority if it has been treated relatively unfair Assign credits to each link to keep track of relative fairness

11. Credit Based Approach (4/5) Credit Schema One credit is charged from its credit account if that link is used If a slot is not used at all a special temporary account is debited To keep the system balanced increase the temporary if a credit is debited from a link’s account Redistribute temp accnt’s savings when it reaches n, which is # of links (QoS charecteristics may be considered)

12. Credit Based Approach (5/5) Credit Schema Abort an ongoing sniff event if link with upcoming sniff slot has a higher credit Poll – Null : Switch to next PP If max. poll interval exceeds give highest priority

13. Performance Enhancements to minimize # of switches After communcation begins reserve some number of slots for that comm May lead to starvation Switch links only if the difference between the credits is higher that a threshold

14. Performance Enhancements for better utilization of Redistribution Redistribute a link’s credits if it doesn’t use them Absolute vs. Relative distribution Relative distribution: Triggered by Poll-Null

15. Performance Enhancements Adaptive Presence Point Density (APPD) Do not waste credit with unseccessful sniff slots or Poll-Null sequences Double PP intervals for unsuccessfull tries Assure process is identical at both sides

16. Conclusion Simulation results are in the paper Good for arbitrary topologies No need for scatternet-wide coordination Easily adapts to the traffic conditions QoS aspects are open problems I am impressed!!!