ACOE 422
Network Design and Planning Issues and Performance Evaluation
Outline
Network DesignRadio Network PlanningPerformance EvaluationCase Study 1:
WLAN Coverage Planning
Case Study 2: WLAN Performance Evaluation
Introduction Wireless networks rely on an inexpensive but
prone to errors medium (air) with limited bandwidth
We require wireless networks to be Functional Affordable Scalable Flexible Manageable Secure Resilient and Reliable Meet the growing user demands (e.g. of bandwidth) Low cost of ownership consistent with these objectives
Network Design and Planning is very essential!
Coverage
Figure 2.13 A predicted coverage plot for three access points in a modern large lecture hall. (Courtesy of Wireless Valley Communications, Inc., ©2000, all rights reserved.)
Figure 2.15 A typical neighborhood where high speed license free WLAN service from the street might be contemplated [Dur98b].
Figure 2.16 Measured values of path loss using a street-mounted lamp-post transmitter at 5.8 GHz, for various types of customer premise antenna [from [Dur98], ©IEEE].
Network Design
Specify network architecture Define radio access network design and engineering Define core network design and engineering Provide detailed protocol design
Traffic Modeling Decide on voice and data applications
Mobility Modeling Mobility assessment and design is important
Complete area plans Provide performance and bottleneck analysis Specify security and redundancy plans
Network Design: considerations
Services & Traffic: How much and where? Impact on network quality, efficiency and cost
What is best design strategy (given an imprecise demand forecast)? Coverage vs. capacity, cell breathing (UMTS) Ability to use existing sites (e.g. GSM)
Meet budget and cash flow constraints
Radio Network Planning RNP includes:
Dimensioning Detailed Coverage & Capacity Planning Network Optimization
Dimensioning estimates: an approximate number of base station sites base stations and their configurations other network elements
based on the operator’s requirements and the radio propagation in the area
Dimensioning must fulfil certain requirements for: Coverage Capacity Quality of Service (QoS)
Radio Network Planning
Coverage and Capacity Planning Determine the coverage regions, area type information
and propagation conditions Determine the available spectrum and traffic density
information Note: In W-CDMA networks (e.g. UMTS),
capacity and coverage are closely related both must be considered simultaneously in the planning
process Network Optimization
Provide optimal coverage probability, blocking probability and end user throughput
Radio Network Planning
Outputs during RNP: Rough number of base stations and sites Base station configuration Site selection Cell specific parameters for RRM & adjusting of
RRM parameters to optimal values Analysis in the issues of capacity, coverage
and QoS
Performance Evaluation
Takes place prior to the deployment of a system
Assesses a system’s capabilitiesEvaluates any new mechanisms the
system will useNote: System = a collection of related
entities that interact together over a time to accomplish a goal E.g. to deliver telecommunication services that
satisfy specific QoS requirements
Performance Evaluation
Two ways to achieve performance evaluation of any system Experiment with the actual system Experiment with a model of the system
Experiment with the actual system Set up the system and run it Collect measurements that will aid in the assessment of
the system Exact results but costly Often the system is not available
Performance Evaluation Experiment with a model of the system
The model can be physical or abstract abstract = representation of the system containing
structural, logical or mathematical relationships The physical model is evaluated similarly to an actual
system The abstract model may be evaluated in two ways:
Analysis (mathematical analysis) Simulation
Mathematical analysis Costly Requires specialized knowledge Often several approximations need to be made (for
complex systems) hard to generalize results Simulation becomes more and more popular
Performance Evaluation
Simulation Simulation models may be categorized
according to the type of input data they accept Deterministic Stochastic
Simulation models may be categorized according to the factors that cause system state to change
Continuous (time-based) Discrete event-based (still requires a time-keeping
mechanism to advance from one event to another)
Performance EvaluationEvaluation of a system
Experiment with the actual system
•Costly
•Often the system is not available
Experiment with a model of the system
Physical model Abstract model
Analytical evaluation (mathematical analysis)
•Costly
•Approximations due to complexity
Simulation
Categorized according to the type of input data it accepts:
•Deterministic
•Stochastic
Categorized according to the factors that cause system state to change
•Continuous
•Discrete Event Based
Conclusions Review your existing applications and infrastructure
Incorporate, as needed, wireless access points, routers, gateways, security devices and middleware
Determine connectivity requirements for your network and mobile devices Integrate seamlessly with your current and future IT
infrastructure Evaluate performance, scalability, and availability metrics
Leverage simulation and modeling tools to help ensure consistent quality of service
Assess server capacity and network coverage Ascertain security and management requirements
Provide maximum security for the whole network infrastructure
Case Study 1: WLAN Coverage Planning
Paper: WLAN Coverage Planning: Optimization Models and Algorithms, E. Amaldi, A. Capone, M. Cesana, F. Malucelli, F. Palazzo
Case Study 1: WLAN Coverage Planning
WLAN medium access mechanism: “listen before talk” approach if a user terminal is covered by more than 1 AP
and is transmitting/receiving to/from one of them, the other APs cannot transmit/receive to/from other users.
causes limited system capacity when coverage areas overlap
Appropriate positioning of APs is crucial
Case Study 1: WLAN Coverage Planning
Simple way to plan coverage consider a set of possible positions of user terminals in
the service area consider a set of AP candidate sites select a subset of sites in which to install APs so as to
guarantee a high enough signal level to all user terminals
Problem Minimizing the number of APs that cover the complete
set of user terminals is an NP-hard task (a.k.a. cardinality set covering problem)
Case Study 1: WLAN Coverage Planning Heuristics are adopted to provide a sub-optimal
solutions Not all such solutions provide acceptable levels
of capacity and QoS Proposed solution:
2 phases: greedy approach & local search The greedy phase starts from an empty solution and
iteratively adds to the current solution the candidate site which maximizes a certain benefit function (calculated for each candidate site)
The local search phase takes as an input the solution provided by the greedy phase. Then the site “neighborhood” is explored for a better solution, using an objective function. The final solution is the one with the highest objective function.
Case Study 1: WLAN Coverage Planning
Conclusions Coverage planning for WLANs is a hard task An optimal solution is NP-hard
A sub-optimal approach is usually taken Proposed approach uses heuristics and is
composed of two phases: the greedy phase and the local search phase
Results show that this approach achieves better overall capacity than the classical approach, which is based on the minimum cardinality set covering problem.
Performance evaluationIEEE 802.11 (XIV)
Unicast data transfer
DIFS
data
ACK
otherstations
receiver
sender
t
data
DIFS
waiting time contention
SIFS
– station has to wait for DIFS before sending datastation has to wait for DIFS before sending data– receivers acknowledge after waiting for a duration of a receivers acknowledge after waiting for a duration of a
Short Inter-Frame Space (SIFS), if the packet was Short Inter-Frame Space (SIFS), if the packet was received correctlyreceived correctly
Masters thesis
http://eeweb.poly.edu/dgoodman/fainberg.pdf
Case Study 2: Performance Evaluation of Wireless LANs
Paper: Enhancements and Performance Evaluation of Wireless Local Area Networks, Jiaqing Song and Ljiljana Trajkovic.
Performance Evaluation is done using the OPNET simulation tool.
Case Study 2: Performance Evaluation of Wireless LANs
Known problems with WLANs WLAN media is error prone (very high BER) Hidden Terminal problem decreases performance Carrier Sensing (for collision detection) is difficult
a station is incapable of listening to its own transmissions
Investigate 3 approaches for improving WLAN performance tuning the physical layer related parameters tuning the IEEE 802.11 parameters using an enhanced link layer (MAC) protocol
Case Study 2: Performance Evaluation of Wireless LANs
OPNET WLAN models WLAN station
IEEE 802.11 WLAN station includes ON/OFF traffic
source includes sink includes WLAN interface includes receiver/
transmitter pair
Case Study 2: Performance Evaluation of Wireless LANs OPNET WLAN models
WLAN workstation workstation with client/server
applications running over TCP/IP and UPD/IP
supports IEEE 802.11 connections at 1Mbps, 2Mbps, 5.5Mbps or 11Mbps (speed is determined by data rate of connecting link)
WLAN server server with applications running over
TCP/IP and UDP/IP supports IEEE 802.11 connections
at 1Mbps, 2Mbps, 5.5Mbps or 11Mbps (speed is determined by data rate of connecting link)
Case Study 2: Performance Evaluation of Wireless LANs
OPNET WLAN models WLAN access point
wireless router Ethernet interface connects the wireless
network to wired networks
Case Study 2: Performance Evaluation of Wireless LANs
Approach 1: tuning the physical layer related parameters Modified OPNET wlan_mac
process to introduce 4 parameters
Slot time SIFS time Minimum contention window Maximum contention window
To enable choose “customized” option for “Physical Characteristics”
Case Study 2: Performance Evaluation of Wireless LANs
Approach 1: tuning the physical layer related parameters Scenario with 2 WLAN stations WLAN stations have no TCP or higher layers, therefore
reflect the performance of MAC layer protocols more accurately
First set of simulations demonstrates the effect of Slot time and Short Inter-frame Space (SIFS) on WLAN performance
Second set of simulations demonstrates the effect of Minimum Contention window on the average media access delay
Case Study 2: Performance Evaluation of Wireless LANs
Approach 1: tuning the physical layer related parameters Simulation set 1 media access delay in first
node is collected media access delay =
queue delay + contention delay
Results: smaller slot time and SIFS decrease the average media access delay improved performance
Case Study 2: Performance Evaluation of Wireless LANs
Approach 1: tuning the physical layer related parameters Simulation set 2 media access delay is
again collected Results: setting Min
contention window to a smaller value (in the case when there are few WLAN stations in the network) decreases media access delay improved performance
Case Study 2: Performance Evaluation of Wireless LANsApproach 2: tuning the IEEE 802.11
parameters A BER generator was developed and
integrated in the wlan_station model Nine simulation scenarios with various
combinations of values for BER and Fragmentation threshold
to demonstrate the effects of the fragmentation threshold
Throughput is collected Throughput represents the rate of data successfully
received by other stations
Case Study 2: Performance Evaluation of Wireless LANs
Approach 2: tuning the IEEE 802.11 parameters Results show that for low BER various fragmentation
threshold have no significant effect on the WLAN performance.
Case Study 2: Performance Evaluation of Wireless LANs
Approach 2: tuning the IEEE 802.11 parameters Results show that for relatively high BER, a small
fragmentation threshold can significantly improve WLAN performance.
Case Study 2: Performance Evaluation of Wireless LANs
Approach 2: tuning the IEEE 802.11 parameters Results show that for relatively low BER, a very small
fragmentation threshold can significantly deteriorate WLAN performance, because of the heavy packet overhead.
Case Study 2: Performance Evaluation of Wireless LANs
Approach 3: using an enhanced link layer (MAC) protocol Adaptive back-off mechanism was examined This mechanism can be implemented on top of
the existing access scheduling protocol and does not introduce additional overhead.
The main idea of the mechanism is to estimate the shared channel by calculating the slot utilization ratio.
High utilization possible collision back-off
Case Study 2: Performance Evaluation of Wireless LANs Approach 3: using an enhanced link layer (MAC)
protocol adaptive back-off mechanism was implemented and
integrated into the wlan_mac process model.
Case Study 2: Performance Evaluation of Wireless LANs
Approach 3: using an enhanced link layer (MAC) protocol Three simulation scenarios with various numbers of
identical WLAN stations Data is sent at an average rate of 820kbps Destination stations are randomly chosen by the source
station Results collected for analysis include:
Throughput (rate of data successfully received by other stations)
Load (rate of data sent to other stations)
Case Study 2: Performance Evaluation of Wireless LANs
Approach 3: using an enhanced link layer (MAC) protocol Results: with the adaptive back-off mechanism load can be greatly
reduced while throughput can still achieve the same or higher value. the mechanism can effectively reduce the number of collisions and data loss
Case Study 2: Performance Evaluation of Wireless LANs Approach 3: using an enhanced link layer (MAC)
protocol Results: throughput/load behavior of WLAN with more nodes is
consistent
Case Study 2: Performance Evaluation of Wireless LANs
Conclusions 3 methods for improving WLAN performance were
implemented in OPNET Tuning the physical layer characteristics can greatly
improve network performance Properly chosen values for fragmentation threshold
improves WLAN performance when BER is high The adaptive back-off algorithm in the MAC layer can
effectively reduce the number of collisions This case study used simulation as the performance
evaluation method and came to its conclusions after a series of simulation sets for different scenarios