1. Submission Title: [Distributed Contention Access Scheme]
Project: IEEE P Working Group for Wireless Personal Area Networks (WPANs)
Submission Title: [Distributed Contention Access Scheme]
Date Submitted: [Sep 5, 2008 ]
Source: [Paul Dixon, Pei Liu, Liang Li ] Company: [Hisilicom Technology Co. Ltd, Vinno Technologies Inc. ]
Address: [Nan Tian Bld., No.10 Xinxi. Rd Shang-Di Information Industry Base, Hai-Dain District, Beijing,China ]
Voice:[ ],
Re: [ IEEE e ]
Abstract: [The requirement to upgrade CSMA performance for busy systems]
Purpose: [To encourage discussion.]
Notice: This document has been prepared to assist the IEEE P It is offered as a basis for discussion and is not binding on the contributing individual(s) or organization(s). The material in this document is subject to change in form and content after further study. The contributor(s) reserve(s) the right to add, amend or withdraw material contained herein.
Release: The contributor acknowledges and accepts that this contribution becomes the property of IEEE and may be made publicly available by P
Submission
Paul Dixon, Pei Liu, Hisilicon.

2. The Features of WPAN Communication
low data rate and small packets from a single node
Large and variable number of nodes
CSMA/CA mechanism: slotted and unslotted
It is necessary to enhance the CAP Scheme to optimise the network throughput
We must consider the feasibility of optimising the CAP Scheme (based on the CAP model and calculation method of throughput)
Submission
Paul Dixon, Pei Liu, Hisilicon.

3. The current scheme works only over a small range of load
As the number of competitors for access to the channel varies, many potential problems arise
There can be large swings in throughput and efficiency
Limits on the number of Backoffs and Retires are easily exceeded
Submission
Paul Dixon, Pei Liu, Hisilicon.

4. A simple assessment of parameters quickly identifies the dangers.
The Current Scheme
A simple assessment of parameters quickly identifies the dangers.
The optimum range of back off values varies with the number of competitors for access
For efficient operation over the full range of loads, back off limits must be variable
Submission
Paul Dixon, Pei Liu, Hisilicon.

5. This count takes a random value between 0 and 2 ^ be - 1
Back Off
The current scheme assigns a random backoff count for attempts to gain access to the channel.
This count takes a random value between 0 and 2 ^ be - 1
According to the default parameters, be starts with a value 3 and increments on each backoff to a maximum value of 5
Submission
Paul Dixon, Pei Liu, Hisilicon.

6. Back Off
The default of the current scheme permits a maximum of 4 Backoffs before a declaring a Channel Access Failure
Simulation demonstrates the fragility of this limit as Channel Access Failures can occur with as few as two competitors if there is an intense burst of activity
Submission
Paul Dixon, Pei Liu, Hisilicon.

7. Using the Backoff Period as the unit of time
Assuming the average random number to be mid value, four backoff periods corresponds to 44 units of time
Typical period of a Transmission is 15 to 25 periods, depending on size
Submission
Paul Dixon, Pei Liu, Hisilicon.

8. However collisions need to be minimised to enhance throughput
In a busy system, collisions reduce the throughput by causing periods of time to be lost during these collisions
In terms of Channel Access Failures, collisions do not seem to as great a problem as Backoffs
However collisions need to be minimised to enhance throughput
Submission
Paul Dixon, Pei Liu, Hisilicon.

9. Conclusion
As the system becomes busier with a larger number of competitors attempting to push the system to the limits of its throughput capacity, a new analysis of the channel access method is required
This must address the assignment of the backoff variable as well as limits placed on the number of Backoffs and Retries
Submission
Paul Dixon, Pei Liu, Hisilicon.

10. Analysis
If we are to cater for n, a larger number of competitors, where the average transmission period is 20 units, then if each device could be made to cycle in sequence, it is clear that the wait of a device for its next turn could be of the order of 20n.
This defines the backoff period that must be considered typical and catered for by the duration of each backoff and limit on the number of Backoffs permitted
Submission
Paul Dixon, Pei Liu, Hisilicon.

11. Analysis
Equally with n competitors, the probability of two devices beginning their CCA sequence at the same time and proceeding to collide is a function of the number od devices, n, and the range of backoff variables available
Submission
Paul Dixon, Pei Liu, Hisilicon.

12. Analysis
Simulation reveals a number of results, some of which are not immediately and intuitively obvious
The number of Backoffs and Retries rises with throughput but is not dependent on the number of devices competing for that throughput
Backoffs up to 50 and Retries up to 10 are typical of an optimised strategy
Submission
Paul Dixon, Pei Liu, Hisilicon.

13. Distributed Contention Access
Distributed CAP is proposed to optimise the CAP scheme for:
The channel model within CAP; And
The calculation method of network throughput; And
The performance of network throughput; And
Efficiency of total energy utilisation of network
Submission
Paul Dixon, Pei Liu, Hisilicon.

14. The Distributed Contention Access Scheme (1)
The model of CAP optimisation
optimisation of network
calculate the back off number
optimisation
get the time parameters
calculate successful CCA probability
get the network parameters
Submission
Paul Dixon, Pei Liu, Hisilicon.

15. The Distributed Contention Access Scheme (2)
The detail CAP optimisation steps：
Get the number of devices in the network using the CAP channel model.
Calculate the back off number to get optimal network throughput.
Revise the device number in CAP period based on the result of 2nd step
Submission
Paul Dixon, Pei Liu, Hisilicon.

16. The Distributed Contention Access Scheme (3)
Procedure of optimal contention access
Get the parameters of a cycle
Calculate throughput and average cycle length
Calculate optimal back off number
Calculate the probability of successful CCA
Estimate the device number based on average successful CCA probability
Revise the device number in the every successful CCA cycle
Optimize the network
Submission
Paul Dixon, Pei Liu, Hisilicon.

17. Simulation Results of the CAP Optimisation (1)
Supposed:
Some devices transmit fixed size packets to the coordinator simultaneously, and
Each device can use the distributed CAP for optimisation.
The initial number of devices is 5, and 80s later, some additional devices are connected into the network
Other conditions strictly follow the ones in CSMA/CA procedure in std
Submission
Paul Dixon, Pei Liu, Hisilicon.

18. Simulation Results of the CAP Optimisation (2)
Case 1: 15 additional devices become active
Submission
Paul Dixon, Pei Liu, Hisilicon.

19. Simulation Results of the CAP Optimisation (3)
Case 2: 35 additional devices become active
Submission
Paul Dixon, Pei Liu, Hisilicon.

20. Simulation Summary
By CAP optimisation using Distributed Contention Access, the throughput of network will reach a new stable optimal value after short-lived fluctuation, even though the device number changes sharply
By the optimisation of CAP and throughput, the efficiency of total energy utilisation of WPAN is improved.
Submission
Paul Dixon, Pei Liu, Hisilicon.