Adaptive QoS Control Based on Benefit Optimization for Video Servers Providing Differentiated Servic

1
Adaptive QoS Control Based on Benefit
Optimization for Video Servers Providing
Differentiated Services

• Aaron Slotness
• Karen Hughes
• Tim Fellman

2
Introduction and Background

• Video server
• Different levels of QoS
• Media types

3
Key Terms

• Quality of Service (QoS)
• Disk scheduling
• Arrival rate (?)
• Departure rate
• Cycle time TSR
• Resource reservation

4
Assumptions

• Multiple video files on a single disk
• Each video 3 different version
• Arrival rate (?) and departure rate (µ)
• Server uses efficient disk scheduling TSR
• Admission control algorithm
• Resource recovery algorithm

5
Assumptions contd

• User stays until service completed
• (Baseline QoS control algorithm)
• Users dont like arbitrary raising or lowering of
  QoS levels
• Users could accept having their QoS level raised,
  but never lowered
• (Greedy Algorithm)

6
Benefit Optimization

• Each QoS level assigned a benefit value
• User pays based on QoS level received
• Usage of penalty value q

7
Baseline QoS control algorithm
8
Greedy QoS control algorithm

• Recall move a user to better QoS level, but not
  worse
• baseline algorithm does not allow for
  redistribution of resources for existing users
• users get QoS elevation based on longest
  connect-time
• QoS elevations continue until all server
  resources are reallocated.

9
Resource Reservation
10
Resource Reservation Cont.

• System queuing
• Partitions
• Priority based?

11
Analysis

• Highest-QoS-only (HQO)
• Lowest-QoS-only (LQO)
• Equal-Share (ES)

12
Numerical data and graphs
13
Numerical data and graphs
14
Numerical data and graphs
15
Numerical data and graphs
16
Numerical data and graphs
17
Numerical data and graphs
18
Numerical data and graphs
19
Numerical data and graphs
20
Numerical data and graphs
21
Reactions and Opinions
22
Questions?