Adaptive resource management with dynamic reallocation for layered multimedia on wireless mobile communication net work

1
Adaptive resource management with dynamic
reallocation for layered multimedia on wireless
mobile communication net work

• Date2005/06/07
• StudentJia-Hao Xu
• AdvisorKai-Wei Ke

2
Outline

• Motivation
• Background
• Multimedia Streaming Service
• Call admission Control for Single-Media
• Call admission Control for Multimedia
• Numerical Results
• Conclusion

3
Motivation

• We want to find an optimal call admission control
  policy for wireless multimedia streaming service.
• Main issues
• To promote Wireless bandwidth resources
  utilization.
• To guarantee handover call blocking probability.

4
Background

• Wireless communication network introduction
• Wireless resource allocation method
• MDP (Markov Decision Process) introduction

5
Wireless communication network introduction
6
Wireless Resource allocation method 2GGSM
(Global System for Mobile)

• It uses Frequency Division Multiple Access (FDMA)
  and Time Division Multiple Access (TDMA) to
  access air interface.
• Divide carrier into multi TDMA frame (4.615ms)
  and each frame consist of eight time slots
  (577us).
• It uses 51-Multiframe (51 TDMA frame).

7
Wireless Resource allocation method 2GGSM
(Global System for Mobile)
The unit of resources allocation is time slot.
8
Wireless Resource allocation method 2.5GGPRS
(General Packet Radio Service)

• The difference between GSM and GPRS is GPRS using
  52-Multiframe (52 TDMA frame).
• There are 4 IDLE frames and 12 Radio Blocks (each
  block contains 4 continuous frame).

9
Wireless Resource allocation method 2.5GGPRS
(General Packet Radio Service)
The unit of resources allocation is radio block.
10
Wireless Resource allocation method 3GWCDMA
(Wideband Code Division Multiple Access)

• It uses Code Division Multiple Access (CDMA) to
  access air interface and Direct Sequence
  Spreading Code (DSSC) modulation technique.
• Each user allocate a orthogonal spreading code
  produced by OVSF Code (Orthogonal Variable
  Spreading Factor Code) technique.
• The transmission depends on Spreading Factor.

11
Wireless Resource allocation method 3GWCDMA
(OVSF Tree)
The unit of resources allocation is SF (4256).
12
MDP introduction

• A Markov Decision Process is just like a Markov
  Process, except the transition depends on the
  action at each time step.
• A systems state has various options (actions) to
  choose from, and each option associates a
  reward.
• The goal is to find a optimal policy, which
  specifies which action to take in each state, so
  as to maximize rewards. (solved by LP)

13
Outline

• Motivation
• Background
• Multimedia Streaming Service
• Call admission Control for Single-Media
• Numerical Results
• Conclusion

14
Streaming

• Streaming is a process of playing a file while it
  is still downloading. And, it uses buffering
  techniques.
• When a large media file (audio, video, etc.) is
  broken into smaller pieces so it can viewed or
  heard immediately. This avoids the wait for the
  whole file to be downloaded first.
• It doesnt require low delay but low jitter and
  media synchronization.

15
Multimedia application

• There are many application using this technique,
  such as video phone, video-on-demand and
  interactive video games.
• These services consume lots of bandwidth
  resources.
• Layered encoding (multirate and adaptive)
• A technique to convert a file into a compressed,
  streaming format. For example, H.263, MPEG-2, and
  MPEG-4

16
Streaming service architecture
17
Our streaming architecture
System Web
Select multimedia file
Terminal user(traveler)
Base station (do some things)
Get file from streaming server
Wireless transmission
Return multimedia file
Streaming server
For example, adaptive and reallocation resources
18
Main issuse

• Design a dynamic and adaptive resource allocation
  strategy to
• Lower Blocking probabilitynew and handover call
  (especially)
• Increase resources utilization

19
Outline

• Motivation
• Background
• Multimedia Streaming Service
• Call admission Control for Single-Media
• Numerical Results
• Conclusion

20
Network model
21
Network model (cont.)

• Each base station has a total of C bandwidth
  resources and support k different types of
  encoding method (means k different rates).
• We assume that new arriving calls in a cell
  follow a Poisson process.
• Both call session time and call dwell time are
  exponentially distributed and are independent
  from cells to cells

22
Network model (cont.)
23
MDP-based Call Admission Control
24
MDP-based Call Admission Control (cont.)
25
MDP-based Call Admission Control (cont.)
26
MDP-based Call Admission Control (cont.)
27
MDP-based Call Admission Control (cont.)
28
Three Strategies for

• Adaptive Only
• Adaptive and Dynamic Reallocation Finally
• Adaptive and Dynamic Reallocation Anytime

29
Adaptive Only (AO) --
X (0,0) and call arrival gt (1,0) or (0,1)
30
Adaptive Only (AO) --

31
Adaptive Only (AO) --
32
Adaptive Only (AO) --

33
Adaptive Only (AO) -- example
34
Adaptive and Dynamic Reallocation Finally (ADRF)
--
X (1,2) gt (0,3)or(1,2)or (2,1)or(3,0)
35
Adaptive and Dynamic Reallocation Finally (ADRF)
--

36
Adaptive and Dynamic Reallocation Finally (ADRF)
--

37
Adaptive and Dynamic Reallocation Finally (ADRF)
--
38
Adaptive and Dynamic Reallocation Finally (ADRF)
--
39
Adaptive and Dynamic Reallocation Finally (ADRF)
-- example
40
Adaptive and Dynamic Reallocation Anytime (ARDA)
--
41
Adaptive and Dynamic Reallocation Anytime (ARDA)
--

42
Adaptive and Dynamic Reallocation Anytime (ARDA)
--
43
Adaptive and Dynamic Reallocation Anytime (ARDA)
--
44
Adaptive and Dynamic Reallocation Anytime (ARDA)
-- example
45
Outline

• Motivation
• Background
• Multimedia Streaming Service
• Call admission Control for Single-Media
• Numerical Results
• Conclusion

46
Numerical results Adjust Offered Load
47
Numerical results Adjust Offered Load
48
Numerical results Adjust Offered Load
49
Numerical results Adjust handover blocking
probability upper bound
50
Numerical results Adjust handover blocking
probability upper bound
51
Numerical results Adjust handover blocking
probability upper bound
52
Numerical results Adjust new and handover call
ratio
53
Numerical results Adjust new and handover call
ratio
54
Numerical results Adjust new and handover call
ratio
55
Why ADRF equals to ADRA ?

• Look static state probability

56
Why ADRF equals to ADRA ? (cont.)

• Look taking which action

57
Conclusion

• We propose three call admission control policy
  AO, ADRF and ADRA
• Through numerical analysis, we find that AO
  method is the worst strategy, ADRF and ADRA are
  the best strategy in single-media streaming
  service.