QOS Issues in Servers for Wireless Communications

1
QOS Issues in Servers for Wireless
Communications

• Dept. of Business Administration, YDU
• Ming-Chung Tang

2
Outline

• Motivation
• Wireless Servers QoS
• On-line Scheduling Overiew
• Adjustable On-line Scheduling Algorithms
• Real-time Applications
• Conclusion
• Future Work

3
Motivation
4
GSM System

• Software Radio
• GSM SMS / Applications
• WAP Gateway
• Real-time / QoS Issues
• Imprecise Computation Model

5
Speech Data in GSM
6
Wireless Servers QoS
7
Wireless Servers

• Servers
• Content Provisioning, Proxy,
• Gateways
• Filtering, Adapting,
• Base Stations
• Message Passing, Location Info,

8
QoS Issues in Wireless Servers

• QoS for which layer?
• QoS over what kind of bearer?
• QoS objective?

9
QoS / Real-time on Internet

• Techniques for Internet Environment
• Int-Serv / Diff-Serv for QoS Provisioning
• RTP / RTCP for Real-time Services

10
Our Research Efforts

• QoS in terms of
• Real-time Online Scheduling
• Employ Imprecise Computation Model
• To Provide Flexibility

11
(No Transcript)
12
On-line Scheduling Overview
13
Real-time Scheduling

• Timing Constraint
• Task Characteristics
• Arrival / Ready / Processing Times, Deadlines
• Scheduling Algorithm
• Feasible Schedule

14
Real-time Systems
15
On-line Scheduling

• QoS Model
• The (m,k)-firm Model
• m out of k consecutive arrival tasks meeting
  their respective deadlines
• The Imprecise Computation Model
• A mandatory a optional portion (subtask) for
  each arrival task
• All mandatory subtasks have to meet their
  respective deadlines

16
An Example
17
Adjustable On-line Scheduling Algorithms
18
On-line Scheduling

• For Imprecise Computation Model
• Traditional Objectives
• Minimize Total Error (e.g. Algorithm NORA)
• Maximize Value Obtained from Scheduling Results
  (For a Firm-deadline System)
• Reduce Task Rejection Rate
• Optimality
• Generally, No Optimal Algorithms for the Problem
• IF under FMC, Optimal Algorithms Exist

19
On-line Scheduling

• Our Objectives
• Extending NORA for Adjustable Schedulability
  QoS
• KeepingSei minimized
• Problem Formulation
• A set of on-line, preemptive, imprecise tasks T
  T1, T2, , Tn
• Each task Ti is characterized by
• ai, ri , di , pi , mi , oi ,,xi , ei

20
Reservation List and NORA
M1
M2
M3
O3
O2
M1
M2
M3
21
The Proposed Algorithms

• MOS, MOP, and MOF
• For different QoS considerations
• Using K-tasks-look-ahead Substitutable Check for
  Oi

22
An Example for MOS (k 1)
M3
M4
M3
O2?
23
An Example for MOS (k 1)
24
An Example for MOP (k 1)
O1
25
An Example for MOF (k 1)
O1?
26
Another Example
27
Optimality of Algorithm MOS

• Algorithm MOS produces the minimum total error
  when the given task set satisfies FMC.
• Algorithm MOS has the same complexity as NORA.
• Algorithm MOS is optimal in the sense that it can
  minimize the total error under FMC.

28
Overview of QSM

• Features of QSM
• Applying Imprecise Computation Model
• Real-time, On-line Scheduling
• Coarse-and-fine adjustable

29
The QSM
30
The Problem of QSM

• The Regulator
• Hard to find a stable one to maximize the value
  of scheduling result
• Scheduling on Faster Machines
• Shortening processing times of tasks
• Applying Competitive Analysis

31
Competitive Analysis

• Competitive Ratio
• s-speed c-approximation
• Value Function for Imprecise Tasks

32
Modified QSM
33
Real-time Applications
34
Last Mile Proxy Server

• QoS Guarantees for Web Contents Delivering at
  Base Station Side
• Real-time
• Multiple QoS Level Management
• Power-aware

35
Q-Bridge
36
Bridging Model

• Applying Imprecise Computation Model
• None-preemptive Tasks for Packets
• Algorithm Comparison
• NORA_np
• MF_np

37
Simulations

• Simulation Conditions
• ri ai . (packets are ready upon arrival)
• The mean packet arrival time (mu1) is fixed at
  50.
• The mean packet service time (mu2) is ranging
  from 40 to 110.
• Each packet set T contains 1000 arrival packets.
• The imprecise ratio for each arrival packet is
  ranging from 0.5 to 0.09.

38
Simulation Results (1/4)
39
Simulation Results (2/4)
40
Simulation Results (3/4)
41
Simulation Results (4/4)
42
Software Radio (SWR)

• Benefits
• Rapid prototyping
• Experimentation
• Cost reduction
• Easily deployed updates
• System Flexibility

43
Ideal Software Radio
44
Component-binding Based SWR
45
Binding CFC with DFCs
46
GSM-based SWR
47
Recent Progress of SWR

• Vanu Inc.
• SWR on PDA
• FM radio / APCO 25
• GNU Radio
• FM radio / HDTV

48
Conclusions

• Research Summary Contributions
• On-line Adjustable Scheduling Algorithms
• QSM Architecture
• QoS-enabled Bridging Model
• Software Radio Architecture

49
Future Work

• Applying Competitive Analysis
• On faster machines in imprecise model
• Extending Q-Bridge
• Roaming / Ad hoc networking

50
Thank You.