Performance Evaluation Study for HiperLan WLAN Protocol
Omar A. Elprince
May 20, 2003
Table of Contents
1. Introduction
2. Hiperlan overview
2.1 Hiperlan Protocol architecture and layout
2.2 HIPERLAN MAC Protocol
3. System Performance
4. The Channel Access Mechanism
4.1 Hiperlan Access Mechanism – EY-NPMA
4.2 ETSI RES10 Parameter Specification
5. Hiperlan Power Saving
6. Hiperlan Simulation
6.1 Simulation Goals and Assumptions
6.2 Simulation Results
6.2.1 Influence on increasing number of mobile stations
6.2.2 Influence on increasing packet size
7. Hidden Terminal Problem......
8. Hiperlan2 versus IEEE 802.11
9. References
Table of Figures
Figure 1. Hiperlan Basic MAC frame Structure Examples [1]
Figure 2. Channel access cycle in the HIPERLAN-1 [8]
Figure 3. Throughput for 2,8,25 stations
Figure 4. Throughput with different packet sizes
Figure 5. Delay with different packet sizes
Abstract
The paper project discusses the analysis and simulation of HIPERLAN MAC protocol, which is a standard protocol for wireless local area networks defined by the European Telecommunications Standards Institute (ETSI). Paper will introduce the HIPERLAN performance problems which affects in a direct and indirect way the system stability.
Moreover, considering the HIPERLAN MAC protocol we will go deep in the structure of the MAC FRAM and selecting the variables that needs to be studied and analyzed thoroughly to form a simulation model and building a simulator in C++ customized framework that can simulate and analyze the system performance.
1. Introduction
Hiperlan network typically has a topology. The Mobile Terminals (MT)communicate with the Access Points (AP) over an air interface as defined by the HiperLAN standard.There is also a direct mode of communication between two MTs, which is still in its early phase ofdevelopment. The user of the MT may movearound freely in the HiperLAN network, which will ensure that the user and the MT get the bestpossibletransmission performance. An MT, after association has been performed (can be viewed as a login), onlycommunicates with one AP in each point in time. The APs see to that the radio network is automaticallyconfigured, taking into account changes in radio network topology.
The general features of the HiperLAN technology can be summarized in the following list.
· High-speed transmission
· Connection-oriented
· Quality-of-Service (QoS) support
· Automatic frequency allocation
· Security support
· Mobility support
· Network & application independent
· Power save
The aim of this paper report is to study the Hiperlan performance and specification in respect to access channel mechanism in the next section I will present the Hiperlan protocols layers, architecture and MAC protocol as an overview. After that I will go deep to the MAC layer protocol and access schemes.
What is Qos? - QoS is a group of necessities for a certain service which is provided by a network to users. At the user level, the important parameter is the user application. The user-level QoS requirements are Criticality (according to QoS based on data transmission and application type) , Cost (either charging on the basis of the usage time of service or the amount of data) and Security(according to confidentiality, integrity, digital signature capability and authentication).
On the technology and network level, the requirements are Bandwidth (refers to speed or data rate to a user), Timeliness (measured through delay time, response time, and delay variation), and Reliability (according to the measured time or frequency).
For wireless, the QoS indicator is the mobility range (according to the size of the geographical area in which the user can receive the service when moving around, and the size of the area covered by a single base station)
2. Hiperlan overview
2.1 Hiperlan Protocol architecture and layout
Theprotocol stack is divided into a control plane part and a user plane part following the semantics of ISDNfunctional partioning; i.e. user plane includes functions for transmission of traffic over establishedconnections, and the control plane includes functions for the control of connection establishment, release,and supervision.
The HIPERLAN protocol has three basic layers; Physical layer (PHY), Data Link Control layer (DLC),and the Convergence layer (CL). At the moment, there is only control plane functionality defined withinDLC.
The PHY, DLC, and the CL are further detailed in later deliverables.
The physical layer of HIPERLAN-1 uses 200 MHZ at 5.15-5.35 GHz, in European Union it is divided into 5 channels and in the United States it is divided into 6 channel and additional 3 channels at 5.725-5.825 GHz bands.[8]
The transmission power is as high as 1 W and the modulation is single carrier GMSK that includes a DFE that consumes electronic power which makes it challenging for the implementation of fallback data rates.
The physical layer of the HIPERLAN-1 codes 416 bits into 496 bits and with maximum of 47 codewords per packet and 450 bit per packet.
Hiperlan-1 was not considered a successful standard by the European Union, in contradiction Hiperlan2 project made a success start aiming at IP and ATM type services at high data rates for indoor and possibly outdoor applications. It supports both connection (integration with the voice-oriented network) and connectionless services unlike Hiperlan1 that support connectionless services only. The cost of supporting both services is the complexity of MAC layer.
2.2 HIPERLAN MAC Protocol
The MAC protocol is the protocol used for access to the medium (the radio link) with the resultingtransmission of data onto that medium. The control is centralised to the AP which inform the MTs atwhich point in time in the MAC frame they are allowed to transmit their data, which adapts according tothe request for resources from each of the MTs.
Figure 1. Hiperlan Basic MAC frame Structure Examples [1]
The air interface is based on time-division duplex (TDD) and dynamic time-division multiple access(TDMA). I.e. the time-slotted structure of the medium allows for simultaneous communication in bothdownlink and uplink within the same time frame, called MAC frame in HiperLAN. Time slots fordownlink and uplink communication are allocated dynamically depending on the need for transmission
resources. The basic MAC frame structure on the air interface has a fixed duration of 2 ms and comprisestransport channels for broadcast control, frame control, access control, downlink (DL) and uplink (UL) data transmission and random access. All data from both AP and the MTs is transmitted indedicated time slots, except for the random access channel where contention for the same time slot isallowed. The duration of broadcast control is fixed whereas the duration of other fields is dynamicallyadapted to the current traffic situation. The MAC frame and the transport channels form the interface between DLC and the physical layer.
3. System Performance
The main problems of the HIPERLAN performance analysis of HIPERLAN MAC protocol is the analytical models have a high degree of complexity.
There for, building a simulator that can evaluate the performance and analyze the performance result and trying to verify the simulation results is the best approach.
The expected performance results from the Hiperlan MAC simulation is the throughput versus the load percentage in the network with respect to the number of stations; Moreover, the average delay versus the percentage load with respect to the packet size.
4. The Channel Access Mechanism
4.1 Hiperlan Access Mechanism – EY-NPMA
HYPERLAN channel access based on channel sensing and contention resolution scheme called Elimination Yield – Non-preemptive Priority Multiple Access (EY-NPMA).
The stations seek access listen to the channel for a certain time period (1700 bit-periods) If the channel is idle it is allowed to start transmitting without any further processing. Each data frame must be explicitly acknowledged by an ACK transmission from the destination node. This reduces protocol overhead under low load condition, however with load higher than 30% the condition criteria is hardly ever fulfilled.
On the other hand, if the channel is busy the full MAC protocol path has to be taken With EY-NPMA (Elimination Yield – Non-preemptive Priority Multiple Access).
Figure 2. Channel access cycle in the HIPERLAN-1 [8]
A channel access cycle with synchronization begins, it consists of three phases: The prioritization phase, the contention phase, the transmission phase. The objective of the prioritization phase is to allow only nodes with the highest channel access priority, among the contending ones, to participate in the next phase. There are five priority slots, each 256 bit-periods long from 0 to 4 and 0 represent the highest priority level. Each node that has a frame with priority level h senses the channel for the first h prioritization slots (priority detection). If the channel is idle during this interval. Then the node transmits a burst in the h+1 slot (priority Assertion) and it is admitted to contention phase, other wire it stops contending and waits for the next channel access cycle.
The contention phase starts immediately after the transmission of the prioritization burst, and it further consists of two phases: the elimination phase and the yield phase.
The elimination phase consists of at most n elimination slots, each 256 bit-periods long, followed by a 256 bit-periods long elimination survival verification slot. Starting from the first elimination slot, each node transmits a burst for a number B, 0 ≤ B ≤ n, of subsequent elimination slots, according to the following truncated geometric probability distribution.[3]
qn / b = n
After the end of the burst transmission, each node senses the channel for the duration of the elimination survival verification slot if the channel is sensed idle the node is admitted to the yield phase, otherwise it drops itself from contention and waits for the next channel access cycle.
The yield phase starts immediately after the end of the elimination survival verification interval and consists of at most m=14 yield slot, each 64 bit periods long. Each node listens to the channel for a number D, 0 ≤ D ≤ m, or yield slots before (if allowed) beginning transmission. D is a r.v with truncated geometric distribution as follows [3]:
Pr{D=d}={ / (1-p) pd / 0 ≤ dmpn / d = m
If the channel is sensed idle during the yield listening interval, the node is allowed to begin the transmission phase. Other wise the node loses contention and waits for the next channel access cycle.
Real time traffic transmission is supported in HYPERLAN by dynamically varying the channel access mechanism priority depending upon the user priority and the packet residual lifetime.[3][8][6]
Important features of the EY-NPMA [5]
No preemption by frames with higher priority after the priority resolution possible.
Hierarchical independence of performance.
Fair contention resolution of frames with the same priority.
Note: refer to table 1 for parameters.
4.2 ETSI RES10 Parameter Specification
The ETSI Hiperlan project defined system architecture and tables for parameter specifications for standardization.
Parameter / ValueChannel Bit Rate (Mbit/sec) / 23.5
Channel Access Mechanism Priority Levels / 5
Maximum number of subseq. Elimination bursts / 12
Probability of bursting in an Elimination slot / 0.5
Maximum number of subseq. Yield listening / 14
Probability of listing in a Yield slot / 0.9
Table1. Operation parameter Settings [3][6]
High User Priority / Law User PriorityNMRL < 10 msec / 0 / 1
10 msec NMRL < 20 msec / 1 / 2
20 msec NMRL < 40 msec / 2 / 3
40 msec NMRL < 80 msec / 3 / 4
80 msec NMRL / 4 / 4
Table2. Computation of Channel Access mechanism priority. [3][6]
Where NMRL (Normalized MPDU Residual Life Time) is the normalization of the number of hops the packet has to travel to reach its final destination.
The user priority is assigned to each packet according to the type of traffic carried to each packet (e.g. voice, data, multimedia stream ... etc) and from this it determines the maximum channel access priority value. How ever, a residual packet lifetime is the time interval within which the transmission of packet must occur before the packet has to be discarded [3] and that’s why multi hop routing is supported within in the Hiperlan standard protocol.
5. Hiperlan Power Saving
In Hiperlan are two types of nodes - forwarders and non-forwarding. Non forwarders only know their direct neighobors (station within radio range) while forwarders know the network topology. The topology information is received and maintained by continuously transmitting and receiving special conrol PDUs and ageing. If a non forwarder node wants to tranmite a packet to a node not within radio range it either addresses next forwarder or broadcast it to all neighbors stations.
Stations that support forwarding are called forwarders. A forwarder consumption of power increases because it has to receive, buffer and forward packets which are sent to one of its clients.[9]
Since forwarders are not mobile and needs to be connected to a power supply, it is possible that a fragmentation situation occurs in which a HIPERLAN is divided into multiple disjoint communication subnets.
Therefore a device has to be turned off in order to save power, this problem is solved by. reducing power consumption without the loss of functionality.
In terms of power consumption, an equalizer is one of the most expensive parts of the receiver the equalizer should only be turned on when the station is the receiver of a packet. Each packet is divided into a low-bit-rate and a high-bit-rate part in order to decide if a station is the destination of a packet without starting the equalizer. Only stations that determine that they could be the receiver turn on their equalizer to receive the high bit rate part of the packet.
The power saving in Hiperlan is based on a contract between p-saver (the station that wants to save power) and the p-supporter (the station that supports p-saver).
The p-supporter has to queue all packets and schedule the transmissions of these packages during the active intervals of the p-saver. Each p-saver could have other p-supporters as well so that it keeps the protocol simple.
Each multicast packet is transmitted only once to avoid the waste of bandwidth.[9]
6. Hiperlan Simulation
6.1 Simulation Goals and Assumptions
The simulation objective is to measure the performance of the Hiperlan access control protocol on an ad-hoc mode environment. Based on the simulative study and results, I will analyze the performance characteristics and discuss the problems of the protocol that has been shown in the simulation results.
For this simulation, I used CSIM18 software [7] libraries under Linux 7 operating system as object oriented simulation tool C++. I build an object class called packet that has all the information necessary for sending and receiving the packets through the mobile nodes. A mobile node is chosen randomly to send a packet and receive a packet, also the packets inter-arrival times used in exponentially distribution in order to be able to differ load conditions [5].
6.2 Simulation Results
6.2.1 Influence on increasing number of mobile stations
While building the simulation it was intended to show the performance that can be expected from the Hiperlan MAC protocol with respect to the number of stations.
Figure 3. Throughput for 2,8,25 stations
The above chart is athree independent simulation results done for 2, 8 and 25 stations accessing the MAC channel with increasing load. We can see that the overall network throughput decreases with the increasing number of sending stations. Note that the packet sizes for the 3 simulation results was the same and the stations was scattered randomly within the network area and the stations was in random movement while transmission or receiving.
6.2.2 Influence on increasing packet size
The simulation of packet sizes is very critical for the evaluation of the network the challenge is that it is not predictable, what is the packet sizes that will be transmitted over the network because it depends on the applications that generates the data which is the most important aspect in evaluating a network QoS. More over, the packet size is important in the interconnection of the WLAN network with other networks. If the packets transmitter from other networks cannot be equally effective over the interconnected network they have to bundled or fragmented which increase the processing overhead and complexity. There for, it is important to adapt the packet size to the state of the channel since the shorter packet has a higher chance to get through successfully without errors over a bad link.
Figure 4. Throughput with different packet sizes
As we can see it is clear from figure 4 that theperformance of the MAC protocol will degrade when the packet size decreases and that is because the cost of accessing with the MAC random access scheme is independent of the size of the payload, there fore the relative overhead increases with the smaller payload.[5]
Figure 5. Delay with different packet sizes
On the other hand it is shown clearly in figure 5 a severe increase for the access delay of large packets with the increase of the load. We can conclude that it is not fare to use types of applications that generate different packet sizes though it will be most effective to use applications that generate similar packet sizes.
7. Hidden Terminal Problem
The AP communicate with the two terminals which are both in the coverage area of the AP but out of coverage area to each other.
There are two major causes for the hidden terminal degradation which are limited antenna range and shadowing.
Usually the AP is larger than the mobile terminals which will increase the negative impacts of the hidden terminal problem, if it is assumed that the coverage area of the AP and the mobile terminals are the same, it is not assured that all the terminals in the coverage area can hear each other.[8]