Transcript Performance Analysis of TETRA and TAPS and

Performance Analysis of TETRA and
TAPS and Implications for Future
Broadband Public Safety
Communication Systems
Workshop on Broadband Wireless Ad-Hoc
Networks and Services,
12th-13th September 2002,
ETSI, Sophia Antipolis, France
Christian Hoymann, Dirk Kuypers, Peter Sievering,
Peter Stuckmann, Bernhard Walke, Bangnan Xu
© 2002 Chair of Communication Networks, Aachen University
[email protected]
Overview
1. Performance Evaluation of TETRA
2. Performance Evaluation of TAPS

Traffic Models
3. Implications for Broadband Public Safety
Communication Systems
©

Centralized vs. Decentralized

Packet-oriented vs. Channel-oriented

Performance
Mesa-Workshop, 12th-13th September 2002
2
Performance Evaluation of TETRA



©
TETRA Release 1 offers circuit switched speech
services and connectionless or connection
oriented data services with data rates at about
4.8 kbit/s
ETSI defined 10 scenarios for the comparison of
TETRA systems
Scenario 10 (public or private network for
airlines ground services, airport security, fire
brigades) defined highest amount of offered
traffic per terminal
Mesa-Workshop, 12th-13th September 2002
3
Scenario 10: Parameters and
Assumptions


2500 users
Speech:



Data:




Short data (100 byte): 20 h-1
Medium data (2 kbyte): 0.5 h-1
4 TETRA cells
Configurations with


©
3.6 calls/h per user (PP and PMP)
Mean duration 20 s
400, 500, 600, 700 and 800 users per cells
6 carrier frequencies, 1 control channel
Mesa-Workshop, 12th-13th September 2002
4
Simulation Tool
Performance
Simulation
Y
Protocols
formally
specified in SDL
SDL - GR
SDL - PR
ADTs
- Random Generator
- Statistical Evaluation
- Simulation Parameter
- other
C++ Code
Generator
C++
SDL Trace
Simulator
SDL2SPEETCL
SPEET = SDL Performance
Evaluation Environment and
Tools
©
X
Compile
+ Link
Graphical Trace
and Debug
MSC
Start
Signal
Task
SPEET Class Library
Mesa-Workshop, 12th-13th September 2002
State
5
Simulation Concept
Protocol stack
for TETRA
implemented in
SDL
Traffic load
generators for
speech, FTP,
HTTP, SMTP...
Propagation
models: error
pattern files
(here: error free
transmission)
->Evaluation of multi-cellular scenarios under consideration of
co-channel interference
©
Mesa-Workshop, 12th-13th September 2002
6
800
400
RACH access delay [s]
complementary distribution function
complementary distribution function
TETRA: Simulation Results
800
400
Connection set up time [s]
Connection set up times < 300 ms are hard to achieve
©
Mesa-Workshop, 12th-13th September 2002
7
Overview
1. Performance Evaluation of TETRA
2. Performance Evaluation of TAPS

Traffic Models
3. Implications for Broadband Public Safety
Communication Systems
©

Centralized vs. Decentralized

Packet-oriented vs. Channel-oriented

Performance
Mesa-Workshop, 12th-13th September 2002
8
TETRA Advanced Packet
Service





©
IP driven requirement for higher-speed
packet data
TAPS is an overlay system and heavily
based on GPRS and EDGE standards for
GSM
Changes introduced are mainly concerned
with the matching of frequency bands
Net bit rates up to 384 kbit/s
Needs new infrastructure and terminals
Mesa-Workshop, 12th-13th September 2002
9
Traffic Models: WWW




HTTP organizes the transfer of HTML
documents
WWW sessions consist of requests for a
number of objects with a certain object
size
Reading time describes user‘s behavior
Parameters for thin clients:


©
2.5 objects per page
3700 byte mean object size
Mesa-Workshop, 12th-13th September 2002
10
Traffic Models: E-mail
Load arises with the transfer of
messages performed by an SMTP
user
 E-mail size is characterized by two
log2-normal distributions plus an
additional fixed quota (300 byte)

80% text-based e-mails
 20% mails with attached files

©
Mesa-Workshop, 12th-13th September 2002
11
Traffic Models: WAP

WAP is a suite of specifications that defines an
architecture framework containing





A WAP session consists of several requests for a
deck
Parameters used:



©
Optimized protocols (WDP, WTP, WSP)
Compact XML-based content representation (WML,
WBXML)
Other mobile-specific features
20 decks per session (geometric distribution)
511 byte mean CONTENT packet size (log2-normal
distribution)
14.1 s mean interval between decks (neg.-exp.)
Mesa-Workshop, 12th-13th September 2002
12
Simulation Concept
Models for
•Mobile Station
•Base Station
•Serving GPRS
Support Node
•Gateway GPRS
Support Node
have been
implemented
©
Mesa-Workshop, 12th-13th September 2002
13
Simulation Parameters and
Assumptions







©
1, 4, 6 and 8 fixed PDCHs
C/I = 12 dB (13.5% BLEP)
Coding Scheme 2 (CS-2)
LLC and RLC/MAC in acknowledged mode for
WWW, e-mail and WAP
Multislot capability is 1 uplink and 4 downlink
slots
TCP/IP header compression in SNDCP
1500 byte maximum IP datagram size for WAP,
552 byte for TCP-based applications
Mesa-Workshop, 12th-13th September 2002
14
Application Response Time [s]
Pure WWW/e-mail and WAP traffic
4PDCH
WWW
e-mail
WAP,
1PDCH
WAP
Number of MS
©
Mean DL IP Throughput per user [kbit/s]
GPRS: Simulation Results (1)
Pure WWW/e-mail and WAP traffic
e-mail
4PDCH
WWW
WAP
Number of MS
Mesa-Workshop, 12th-13th September 2002
15
Application Response Time [s]
Traffic Mix
WAP
WWW
E-mail
Number of MS
©
Mean DL IP Throughput per user [kbit/s]
GPRS: Simulation Results (2)
Traffic Mix
E-mail
WWW
WAP
Number of MS
Mesa-Workshop, 12th-13th September 2002
16
Overview
1. Performance Evaluation of TETRA
2. Performance Evaluation of TAPS

Traffic Models
3. Implications for Broadband Public Safety
Communication Systems
©

Centralized vs. Decentralized

Packet-oriented vs. Channel-oriented

Performance
Mesa-Workshop, 12th-13th September 2002
17
Implications for Broadband Public
Safety Communication Systems

Self-organizing networks can work
without infrastructure and can be rapidly
deployed. Especially beneficial for




©
temporary application scenarios
extension of radio coverage of fixed
infrastructure radio networks
disaster relief
Robust network as departure and failure
of nodes will not cause a failure of the
whole network
Mesa-Workshop, 12th-13th September 2002
18
Centralized vs. Decentralized



Self-organizing seems to mean that no central
control will be needed.
Provision of Quality of Service (QoS)
requirements may be easily realized by a central
controller.
Centralized solutions suffer from




©
Increased hardware requirements for central controller
in a broadband wireless network
Temporary chaos caused by failure or departure of the
selected central controller
Direct Mode and multihop communication can not be
realized efficiently
Neighboring central controllers must use different
frequencies => Dynamic channel allocation not easy
Mesa-Workshop, 12th-13th September 2002
19
Packet-oriented vs. Channel-oriented


©
The Distributed Coordination Function
(DCF) of IEEE 802.11 is fully
decentralized and self-organizing, but can
not guarantee QoS.
Provision of QoS requirements of high
performance multimedia applications in a
packet-oriented self-organizing wireless
networks appears to be impossible.
Mesa-Workshop, 12th-13th September 2002
20
Wireless Channel-oriented Ad-hoc
Multihop Broadband Network

Inspired from GPRS and DECT concepts:





©
From GPRS: statistical multiplexing
From DECT: dynamic channel selection
Ability to operate in a fully distributed and
efficient manner
Meets QoS demands for different services
Transmission of packets is channeloriented
Mesa-Workshop, 12th-13th September 2002
21
The Hidden Station Problem



Hidden Station: can not sense transmission of sending
WT, but causes interference to the receiving WT, if it
transmits.
Hidden Stations may degrade the network
performance substantially.
Solutions


©
Busy Tone: sent by the receiving WT to make hidden station
aware of ongoing transmission and prevent it from interfering.
A separate narrow band channel and additional hardware is
needed.
RTS/CTS: RTS sent by sending WT. Receiving WT answers with
CTS. WTs that receive RTS and/or CTS deffer their access
according to transmission duration information in RTS/CTS
packets. Some cases remain where due to interference hidden
stations can not receive the CTS packet.
Mesa-Workshop, 12th-13th September 2002
22
W-CHAMB: E-signal
Solving the hidden station problem completely
through transmission of E-signals in minislots.
VBR Packet dropping probability

Scenario: 20 WTs
No E-Signal
PER=3%
With E-Signal
Traffic load
©
Mix of 50% ABR and
50% VBR traffic
Connectivity=0.58
No E-signal:
RTS/CTS mechanism
Minislot length: 10%
of normal slot length
Mesa-Workshop, 12th-13th September 2002
23
W-CHAMB: Network Connectivity

Connectivity=Mean number of neighbors,
normalized by the maximum number
Throughput [%]
C=0.93
C=0.60
C=0.24
Throughput increases
linearly with traffic
load until saturation
Packet dropped, if
maximum delay of
300 ms is exceeded
Smaller connectivity
reduces system
throughput
Length of
connections (hop
count) is reduced
Traffic load
©
Mesa-Workshop, 12th-13th September 2002
24
Conclusions




©
Traffic performance of existing TETRA and
(E)GPRS systems give lower bounds for
achievable delays and throughput in broadband
communication systems
Channel oriented packet transmission is
appropriate to control QoS in a self-organizing
wireless network
A network with decentralized control is best
suited for the operation of a self-organizing
wireless network
Performance analysis by simulation gives input
in early stages of standardization
Mesa-Workshop, 12th-13th September 2002
25
Questions?
Performance Analysis of
TETRA and TAPS and
Implications for Future
Broadband Public Safety
Communications Systems
Chair of Communication Networks, Aachen University
©
Mesa-Workshop, 12th-13th September 2002
26