Scheduling for BroadBand Wireless Access with Adaptive Burst Profiles

1
Scheduling for Broad-Band Wireless Access with
Adaptive Burst Profiles

• Buchholcz Gergo, Dung D. Luong, Tien Van Do
• Budapest University of Technology and Economics

2
What is it about?

• High Altitude Platforms with antennas attached
• Internet access for everyone
• Low-cost solution (compared to satelite systems)

3
Introduction

• Several scheduling algorithms for wireless
  environments
• High bit error rate
• Based on simple channel models (on-off model)
• Do not work in multi-state channel environments
• Bandwidth allocation for users
• Adaptive burts profile
• Users share the same channel
• The BS assigns time slots to each user
• Depending on the channel conditions the BS
  assigns different burst profiles for each user
• Different bit rates same baud rate

4
Motivation

• A new scheduling algorithm for 802.16
• Concentrate on the download direction
• Adaptive burst profiles provide new opportunities
• Better link, network utilization
• Higher throughput
• Effective packet scheduling algorithm
• To provide QoS
• Fairness
• High resource utilization
• Complex channel model (!)

5
Algorithm I.

• The objective of the algorithm
• Decrease the impact of transient channel
  degradation
• Enhance overall throughput
• Guarantee the long term fairness
• Download direction
• Works on HAP
• Takes advantage onadaptive burst profiles

6
Algortihm II.

• Fundamental idea
• Efficient allocation of free time slots to
  improve performance
• The minimum rate of users must be satisfied
• The unfair allocation of the remaining slots may
  improve the systems performance
• Users with good burst profile receive reward
  slots at the expanse of users with bad burst
  profile
• Unfairness must be controlled
• The effect of fading and transient channel
  degradation on system performance can be reduced!

7
Algorithm III.

• Step 1, share the slots to satisfy the guaranteed
  bandwidth for the SSs
• SSs with no data to receive get no slots
• Step 2, share the remaining slots among the SSs
• SSs with better burst profile receive more slots
  according to the weight function
• The value of p can controll the unfairness
• Degree of fairness

8
Algorithm IV.

• The impact of the p parameter on the fairness
• If p is 0 the algorithm operates as Round Robin
• If p is 1 the algorithm shares free slots
  proportional to the burst profile
• For higher values of p the unfairness increases
• If all users have the same profile they share the
  slots in a fair way!

9
Simulation Environment I.

• NS-2.28
• Topology
• Traffic
• FTP
• Different guaranteed bandwidth values (400kBps,
  600kBps, 800kBps)
• Different values of p (0,1.0,2.0)

10
Simulation environment II.

• Channel-condition model
• Markov chain with 6 states
• Different burst profiles for every state
• Modeling a Rayleigh fading channel

H.S. Wang and N. Moayeri. Modeling, Capacity, and
Joint Source/Channel Coding for Rayleigh Fading
Channels. In Proc. of 43rd IEEE Vehicular
Technology Conference, pp. 473479, 1993
11
Results I.

• The average throughput and improvement of APS and
  Round Robin versus p
• For lower guaranteed bandwidth the throughput
  improvement is higher
• If p is equal to 0 the algorithm behaves like
  Round Robin

12
Results II.

• The average throughput of APS and Round Robin
  versus the minimal guaranteed bandwidth
• For higher p values the throughputmprovement
  increases

13
Results III.

• The fairness index versus time
• Fairness converges to 1 in all the cases

14
Conclusions

• We developed a new algorithm
• Improves throughput
• Provide longterm fairness
• Guarantees minimum bandwidth
• Flexible configuration
• Future work
• Improved channel models
• The problem of TCP mice
• High network utilization or low RTT!

15
Thank you!