Performability Modeling in Wireless Mobile Communication Systems

1
Performability Modeling in Wireless Mobile
Communication Systems
Presentation _at_ Verizon Lab
Yonghuan Cao Advisor Dr. Kishor S. Trivedi

• Center for Advanced Computing and Communication
• Department of Electrical and Computer Engineering
• Duke University
• Durham, NC 27705
• Email ycao_at_ee.duke.edu
• Homepage www.ee.duke.edu/ycao

2
Agenda

• Introduction
• Overview of wireless mobile systems
• Why performability modeling?
• Two topics
• Performability of cellular systems w/ control
  channel failure
• Uplink performance in GPRS under bursty data
  traffic
• A short conclusion

3
Challenges in Wireless Mobile

• Limited radio spectrum
• Huge demand but so limited bandwidth
• Error-prone radio link
• Fading signal, less reliable link
• High mobility
• Mobility management complicated (distributed Loc.
  DBs)
• Wide-area wire-line networks to support
• Service diversity
• Traditional voice data, different QoS
  requirements
• Interconnection to the PSTN Internet etc.

4
Wireless -abilities
The measure of the networks ability to perform
designated functions
R.A.S.-ability concerns grow. High-R.A.S. not
only a sale point for equipment vendors and
service providers. Regulatory outage report
required by FCC for public switched telephone
networks (PSTN) may soon apply to wireless.
5
Causes of Service Degradation
Long waiting-time Time-out Service blocking
Resource limit
Resource full
Equipment failures Planned outages (e.g.
upgrade) Human-errors in opr.
Resource loss
Service Incompletion Loss of information

• In a wireless system, the RF resource is the most
  precious resource.
• However, with the rapid growth of data services,
  buffer, bandwidth and routing capability in core
  networks will become more and more important.

6
The Need of Performability Modeling

• New technologies, services standards needs new
  models
• Traditional performance model may not be
  applicable without proper treatment
• Pure performance modeling too optimistic!
• Outage-and-recovery behavior not considered

Performability modeling Performance
Availability Performability A more complete and
balanced picture Both steady-state and transient
solutions are informative
7
Topic - 1
The use of control channel in cellular system is
essential. How to minimize the impact of control
channels failures? How much performance
difference can a protection switch scheme make?

• Performability Modeling and Optimization of
  Cellular Systems with Control Channel Failure and
  Automatic Protection Switch (APS)

• Y. Cao, H.-R. Sun and K. S. Trivedi,
  Performability Analysis of TDMA Cellular Systems,
  PQNet2000, Japan, Nov., 2000.
• H.-R. Sun, Y. Cao, K. S. Trivedi and J. J. Han,
  Availability and performance evaluation for
  automatic protection switching in TDMA wireless
  system, PRDC99, pp15--22, Dec., 1999

8
A TDMA Cellular System

• Each cell has Nb base repeaters (BR)
• Each BR provides M TDM channels
• One control channel resides in one of the BRs

9
Traffic In a Cell
Common Channel Pool
A Cell
10
Performance Measures

• New call blocking probability, Pb
• Percentage of new calls rejected
• Handoff call dropping probability, Pd
• Percentage of calls forcefully terminated
  crossing cells
• Channel utilization, Uc
• Fraction of time in which available channel
  resource is in use

Pb, Pd, and Uc are determined not only by system
parameters (such as no. of channels, call
admission control scheme, etc.), but also
incoming traffic characteristics and call
duration distributions.
11
Guard Channel Scheme

• Handoff dropping less desirable than new call
  blocking!

12
Failures in System

• Platform_down
• The controller or the local area network
  connecting the base repeaters and controller
  going down causing the system as a whole to go
  down.
• Control_down
• The base repeater where the control channel
  resides going down causing the system as a whole
  to go down.
• Base_repeater_down
• Any other base repeater where the control channel
  does not reside going down does not cause the
  system as a whole to go down, but system is
  degraded (partially down).

13
Automatic Protection Switching

• Upon control_down, the failed control channel is
  automatically switched to a channel on a working
  base repeater.

14
Model of System w/o APS
CTMC State (b,k) bNo. of BR up kNo. of
talking channels
15
Model of System w/ APS
CTMC State (b,k) bNo. of BR up kNo. of
talking channels
A Segment of the Composite Markov Chain Model
16
Performability Indices
System Unavailability
Overall New Call Blocking Prob.
Overall Handoff Call Dropping Prob.
17
Hierarchical Decomposition

• Parameters on different time scales ?
  stiffness
• Mean-time-to-failure months
• Recovery minutes or hours
• Call completion time minutes
• Call inter-arrival time seconds
• Hierarchical decomposition
• Numerically well-behaved, less time-consuming
• Good approximation

18
2-level Decomposition

• High level availability
• Failure/recovery of base repeaters, and platform
  of system
• Low level performance
• New call blocking probability and handoff call
  dropping probability for a given number of
  working base repeaters
• Combine together
• Lower level performance measures as reward rates
  assigned to states on high level.

19
Hierarchical Decomposition
High level - availability Failure/recovery of
base repeaters, and platform of system
0,Nb
0,0
0,b
1,Nb
1,0
1,b
20
Reward rate Assignment
State of High Level Availability Model
21
Accuracy of Hierarch. Model
Comparison of composite and hierarchical models
22
Numerical Results (1)
New Call Blocking Probability Improvement by APS
Unavailability in new call blocking probability
23
Numerical Results (2)
Handoff Call Blocking Probability Improvement
by APS
Unavailability in handoff call dropping
probability
24
Topic - 2
GPRS, a 2.5G system. Capacity-on-demand is the
key idea of radio resource sharing among
circuit-switch voice and packet-switched data
users. On packet level, what is the blocking
probability and delay and whats the cause?

• Packet-level Performance Analysis of ALOHA
  Reservation-based MAC in GPRS under Bursty Data
  Traffic

Y. Cao, H.-R. Sun and K. S. Trivedi, Performance
Analysis of Reservation-based Media Access
Protocol with Access Queue and Serving Queue
under Bursty Traffic in GPRS/EGPRS, Wireless
Network (in review), January, 2001.
25
Background

• GPRS, a 2.5G system, to evolve todays TDMA-based
  GSM and tdmaOne towards 3G.
• GSM/tdmaOne are voice-oriented. GPRS supports IP
  or X.25-based data services.
• Circuit-switched voice and packet-switched data
  services coexist. Voice has higher priority.
• Capacity-on-demand concept and multi-slot
  capability. Theoretical data rate up to 172 kbps.
• Targeted data services small-volume data
  applications in the early stage - email, ftp, web
  browsing, stock/weather broadcasting etc. More
  multimedia later (in EDGE etc).

26
GPRS/GSM Architecture
PSTN
MSC/VLR
HLR
BSC
SS7
BTS
EIR
MS
27
Channel in GSM

• Frequency Division/Time Division (FD/TDMA)
• Uplink and downlink have different frequency
  bands (carriers).
• Each carrier divided into 8 slots (bursts) 1
  frame 4.615ms.
• A physical channel ? the repeat of a TDMA slot
  over time.
• A packet data channel (PDCH) is a physical
  channel for data.
• Effective data rate is determined by Coding
  Schemes (CS1-4)
• Multi-frame structures used to accommodate spec.
  channels.

Frequency
Uplink
Frame 1
Downlink
Frame 0
Time slots in a frame
28
Protocol Stack Segmentation
Application
IP/X.25
SNDCP
LLC
RLC
MAC
GSM RF
Mobile terminal
29
Uplink Data Transfer

• A Temporary Block Flow (TBF) needs to be
  established
• MAC a Slotted-ALOHA Reservation Protocol

Hi, BSC, I have data to send. I need to reserve
3 channels to send 1000 bytes.
30
ALOHA Contention

• Collision may happen in the same contention slot

• Capture may be deployed to improve the
  probability of successful contention.
• A simple contention model used Brashe Walke
  97
• Capture probability K(n) Pr(one contention is
  successfully
• Received by BSC given n MS contend in the same
  slot).

31
BSC ACK

• BSC Acknowledgements after successful contention

Mobile, I dont have the channels you asked, but
your request is received in the access queue now.
Dont contend any more. I will notify you when
channels are available.
Mobile, I have the channels you requested. You
are in the serving queue now. Send data now on
the channels assigned.
32
Access Queue (AQ)

• Access Queue successful requests are queued in
  BSC (FIFO)
• What for? To reduce the chance of collisions.
• Potential problem long delay if channels in use
  not released by others mobiles.

Removed When Requested channels become available
Successful request
Access Queue
33
Serving Queue (SQ)

• Serving Queue
• A frame has 8 slots (channels). What if a request
  for 3 channels, but there is only 2 channels. The
  2 channels wasted.
• A round-robin queue with w/ parameters
• Maximum numbers of slots in a round (NM), and
• Maximum numbers of mobiles to accommodate (NMS)
• How a mobile knows which channels assigned to it?
  Broadcast.

Round 1
Round 2
0
1
2
3
4
5
6
7
0
1
2
3
4
5
6
7
0
1
2
3
4
5
6
7
0
User 1
User 2
User3
User 1
User 2
User3
User 1
User 2
NM 12, NMS 3
34
A Data Mobile

• Finite buffer (W bursts)
• Traffic arrives in the units of LLC frames

35
Traffic Model

• LLC frame arrival bursty in nature
• Can be modeled by an On-off model
• a 2-state Markov Modulated Poisson Process (MMPP)

OFF
ON
ON

• The size of a frame (no. of bursts) may also
  vary.
• Probability mass function (pmf) of frame size
• bk Prframe size k

36
Why Analytical Model?

• Previous studies mainly based on discrete event
  simulation (DES). To achieve tight confidence
  interval needs long run times.
• Analytical models are more efficient in terms of
  running time and may shed more light in
  understanding the system.
• Shortcoming of analytical model sometimes
  oversimplified.

37
How to Model

• To avoid oversimplification, we build a
  comprehensive model to reflect the realistic
  system. A complex model is hard to build without
  automated aid.
• The system is time-slotted operating on
  discrete-time basis. However, a GSM frame (the
  time unit) is fine enough in comparison to
  parameters and measures of interest. So, a
  continuous-time model is adopted.
• Still state space too large, a fixed point scheme
  based on hierarchical decomposition is used.
• We use the automated stochastic Petri net tool,
  SPNP by Duke CACC, to specify and solve the
  model. Results are checked with discrete event
  simulation (DES).

38
The SRN Model
LLC arrival on-off
Finite buffer Connection
pmf of LLC frame size
The tagged mobile
The rest (N-1) mobiles
39
Performance Indices

• LLC frame blocking probability, PbLLC
• MMPP arrival ? PASTA no long applies!
• LLC frame delay, tLLC
• Access delay (if no connection established)
• Contention delay
• Waiting time in access queue
• Transmission delay
• Time to transfer the queued bursts before this
  frame
• Time to transfer this frame
• System throughput, ThSYS

40
Data Traffic Parameters

• OFF and ON arrival rates fixed l00.01, l1.99.
• Average OFF and ON periods, (a0-1 and a1-1), are
  measures of burstiness.

If average arrival rate, l0, and l1 are fixed,
the burstiness measure, the index of dispersion
for counts (IDC) Gusella 91, is proportional to
a0-1 or a1-1.

• Three pmfs of LLC frame size are used.

41
Model Accuracy
Simulation 95 CI Written in C
SRN Model Using SPNP
42
Result-1 Effect of Traffic Profile

• Implication to network design
• Segmenting network layer PDU to finer LLC frames
  tends to improve performance.
• But note that overhead may be more for finer
  segmentation. A trade-off.

43
Result-2 Effect of Traffic Burstiness

• The burstier the traffic is, the more likely a
  frame is blocked, the shorter frame delay and the
  less throughput. Traffic shaper may be useful.

44
Result-3 Effect of Request Size

• Multi-slot (multi-channel) improves overall
  performance.

45
Result-4 Components of Frame Delay

• Waiting time in access queue dominates delay (due
  to limited channel).
• Contention delay negligible due to AQ and
  capture.

46
Conclusion

• The continuous-time model captures major features
  in GPRS uplink, able to efficiently provide
  accurate results.
• Interesting properties are revealed in numerical
  results, that are practically useful in network
  design and implementation.
• Results are the base for future model
  simplification.
• Modeling methodology (fixed-point hierarchical
  decomposition) used can be applied to a broad
  range of applications (e.g. CDMA CPCH etc.).

47
The End Thank you!