WPCq 2BJ ZCourier3|WBoldTimes Roman@KX@Apple LaserWriter PlusAPLASPLU.PRSx6X@Khhhh gX@2,<23|WCourierCourier Bold^2CRddCCCdq2C28dddddddddd88qqqYzoCNzoozzC8C^dCYdYdYCdd88d8ddddCN8ddddY`(`lC2CC!CCCCCCCCCCd8YYYYYYzYzYzYzYC8C8C8C8ddddddddddYddddodYYYYYYdzYzYzYzYddddddCdCdCCCdNCdz8zCzCz8dddddCCCoNoNoNoNzCzCdddddzYzYNF2[dCYddddd7>d<d<$YYdCCddooCYd<d<$YYdCCddooCY SIR, while the other packet separation < SIR.(#` .` ` No packets lost when destination addresses are different and both meet separation > SIR condition.(#` XCochannel Interference calculated based on Path attenuation and SIR separation requirement.(# XA Normal Distribution "Fading Margin" uncertainty can be specified and will be applied for all Path attenuations calculations. This means that path attenuation is different for every packet.(# XApart from the errors occuring on the PHY, a uniform distribution error probability can be specified for packet transmissions, and is applied independent of path attenuation.(# XA Micro Wave oven interferer (jammer) can be specified, which uses a programmable onoff duty cycle at a programmable level.(# XA Adjacent Channel environment can be selected that allows both networks to use different frequency bands with the isolation between the channels controlled by a parameter. Note that stations do not defer when inband signal level caused by the other band is above the carrier detect threshold.(# XServer and Workstation processing delay can be specified separately, so that traffic load can be controlled. The packet generation delay is the sum of a given fixed delay and a random delay. All packets are generated independently. The packet arrival process on the medium will therefore constitute a Poisson distribution.(# XAll CSMA/CA parameters and PHY parameters like transmit level and Carrier Sense level can be controlled.(# X"Novell retry timeout" currently limited up to 320 msec, to allow event scheduler to work"0*## in fixed point.(# XAll kind of tallies are maintained that monitor the packet transmission conditions and reason of packet loss. This is done separately for the link to and from the Server.(# The average MSDU delay is calculated for the individual station. XPerformance measured in KBytes/sec actual data throughput excluding Novell overhead in the Perform3 test.(# XPeer to peer performance measured including overhead, and without NCP handshaking. Traffic destinations are random. (# XVery high performance Simulator is "Event driven" and runs about 10 sec per one second simulation on a PC486/33 MHz for 2 fully loaded networks. Event resolution used is 10 sec.(# XProduces two different output files:(# .` ` Detailed report showing performance and lost packet statistics of individual Stations separated in "To the Server" and "From the Server" directions in each network.(#` .` ` Summary report showing throughput and Collision probability per network as a function of an iteration parameter like "Distance between Networks".(#` The Simulation Engine The Simulation Engine is an event driven machine, which will fire off different processes based on events on an event calendar, which are scheduled by the different processes defined in the machine as illustrated in the blockdiagram of fig1. Two Networks can be specified with each consisting of N workstations and one Server (only active in Perform3 test). Packets generated by the Workstations Tx processes are put into a Server RxQueue per network. A separate Server process per network will fetch the packet from the RxQueue, process it for a specified fixed + random Server Processing time and puts a response packet in the Server TxQueue. The Server Tx process per network will fetch its Packets from the TxQueue and transmits it on the Medium. When a process is activated by the process scheduler, it performs the operation relevant for that"0*## state of the process. It will schedule the next state of the process, after a delay time that is put on the event calendar for that process. The Scheduler decrements all events on the event calendar and activates the three different processes when appropriate. This is done on a sampling time interval of 10 sec. So all events are scheduled in units of 10 sec each. A higher resolution can be selected by changing one parameter, at the expense of lower simulator performance. But in the 2 Mbps system under test, 10 sec seems a good tradeoff between accuracy and simulation speed. A separate Medium Manager updates the medium busy status per network on a sample interval bases, and maintains a Medium Busy length counter. Separate handshake signals (SRFlag) are used to notify the scheduler that new packets are put into the Server RxQueue, so that it can activate Server processes when needed. In turn the Server processes can activate a handshake signal (STFlag) to start of a Server Tx process when not already active. When Workstations have transmitted their packet they set up a Timer. This event timer will be reset by the Server Tx process only when it has successfully transmitted a response packet to that workstation. The Queues do only contain workstation addresses, and packet length information. Not shown in Fig1 but in Fig2 and Fig3 are handshaking signals between the processes and the Medium Manager. As can be seen only Tx processes are modelled. For those protocols that only require receiver activity when they are specifically addressed, no receiver process modeling is needed. All evaluation whether the packet was successfully received at the receiver location, is done within the transmit process. Tallies are maintained that monitor the successful and lost packet status, and the reason for packet loss. Workstation TxN Processes The Workstation processes consist of a CSMA/CA State Machine as shown in fig2. In total 7 states are shown. A short description per state follows: Idle:XX` ` X "In this state a packet is scheduled for transmission after a fixed plus random Workstation processing time. For Perform3 (Read mode) only short 64 Byte (Novell request) packets are sent to the Server. For IPXLoad a random mixture of 60% Long and 40% Short packets are scheduled to be sent to a random destination.(# Sense Carrier: "Calculates signal levels of all other transmitters on either of the two networks. When above the CRS Threshold then control is given to the Defer+Backoff state after the Medium busy length as maintained by the"0*## Medium Manager.(# Send PHY preamble:` ` X "In this state the transmitter is actually turnedon after the carrier sense delay time of one slot interval. In this state it is also registered whether a Collision has occurred, and whether capture takes place. In addition it is determined whether there is mutual interference of this packet with possible traffic on the other network.(# Send Data:` ` X "Only schedules the length of the Data part. (Could be combined with other state.(# IFS + Gap:` ` X "At this point the packet is actually transmitted. When the transmision is successful (No collision or jam), then the packet is put in the RXQueue and the SRFlag is set. The transmitter is turned off and the appropriate tallies are updated. The next event is scheduled after the IFS + Gap wait time.(# XX` ` X "Note that the RxQueue is not filled during an IPXLoad test.(# Wait Timeout: "This state schedules the next event after the Novell Timeout timer. It takes this time before the next packet is generated. In case no response was received from the server. This state is not used during an IPXLoad test (no timeout).(# Defer+Backoff: "This state calculates the Backoff period and maintains the backoff counters. When the maximum retry limit is exceeded, then the packet is dropped and counted as lost and the next activity is scheduled after the Novell Timeout period.(# A lot of tallies are maintained by the transmit processes, which are used to update a detailed report on the screen during the simulation. When the test is completed (usually 5 sec real time is simulated), the final result together with the parameter settings of the test, are also stored in a detailed report file. Several flags are used to communicate error conditions between the different transmitters. The actual Medium Busy status is maintained by the Medium manager, which assures that this all happens on the same time, so that it is not related to the order of sequential processing of those events that are scheduled at the same time interval. Also the source and destination addresses of the pending packets are maintained. 0*## Server TxProcess The Server TxProcess works according to the Server CSMA/CA State Machine as shown in fig3. The states are roughly the same as for the workstation. The differences are in the "Idle" and "IFS + Gap" states. The Idle state fetches the destination address for the next packet to transmit from the RxQueue and generates Long response packets that includes the specified data length and the Novell and MAC overhead. The main difference in the IFS + Gap state is that there is no timeout after a packet is transmitted. Also to signal the proper transmission of the packet to the destination workstation its event counter (which was preset to the Novell timeout) is reset so that the appropriate workstation process is activated again (via the Scheduler). Note that the Server TxProcess is not active during the IPXLoad test. Server Process The Server Process is not shown but it is a very simple State machine with 2 states; Idle and Active. It fetches a packet from the RxQueue and puts it in the TxQueue after a fixed + random Server processing time. The momentary status of both the Queues are monitored and reported in the detailed test report. Note that the Server Process is never active during an IPXLoad test. Fading effect modeling discussion The signal path attenuation between any station to each other station is calculated from the distance between the individual stations, and results in an average signal strength at that location determined by the attenuation coefficient parameter of the environment. However, due to fading the actual level can be different from the average level with a probability function according to the Raleigh fading model. When antenna diversity is used as in the WAVELAN product, the probability function shape is becoming close to a normal distribution function. In addition the shadowing effect will be a normal distribution function. To simulate fading effects, a normal distribution fading component is added to every signal path calculation as done in the model. So signal path attenuation between stations for one packet is uncorrelated to the path attenuation of the next packet. This corresponds to a situation where the receive antenna is continuously in motion around the average distance. For the modeled CSMA/CA protocol the independent signal paths as shown in Fig4 are evaluated to determine the access and packet transfer success. Before Network Access, the medium activity is checked. When another transmitter (TxB) is active then it is checked whether the receive level at TxA1 is above or below the Carrier Sense" 0*## threshold. When it is below the threshold then the transmitter will be turned on. Then to determine the transmission success, the wanted signal level at RxA is determined, as well as the interference level at that point generated by the other active transmitter TxB. The difference between both levels must be larger than the SIR parameter for a successful transmission. When at the same time interval a collision occurs with for instance TxA2, then it is determined which of the two signals will be captured by the RxA receiver. Then it is determined which of the packets will be successfully received by RxA, if any (assuming that both TxA1 and TxA2 have the same destination, for instance RxA). Also when TxA1 is turned on, then the effect of the already ongoing transmission between for instance TxB and RxB is determined, by again calculating the signal to interference ratio at location RxB. Again every signal path calculation is the average level at that distance plus a uncorrelated normal distribution "fading component". This covers for instance the situation, that the TxA1 to TxB signal is weak while the TxB to RxA signal can be strong, introducing a high probability of failure at location RxA. The current model evaluates the conditions as shown in figure4. It does not take into account the second order effects, like the effect of multiple interferers. For instance, the interference level at location RxA, is in the described case with also TxA2 and TxB active, the sum of both interference levels, but the model will first evaluate the Capturing, and then evaluate the interference level experienced from TxB. It is not expected that this simplification of the model has any measurable influence. Adjacent Channel model A similar approach is followed for the Adjacent Channel mode. However in this mode, the Carrier Sense function does not "see" the other network, while the interference level is decreased by a "network isolation" parameter. This should account for the side lobe attenuation of the transmit spectrum, and the effect of the receive filter at the other channel. Another effect which is not included in the current model, but which is planned, is the non linear distortion that is generated in the receiver when a strong nearby channel signal causes signal clamping in the receiver. Modeling simplifications Currently the model uses an abrupt boundary to determine the effect of interference and capturing. There is a sharp performance edge at the SIR boundary. In reality there will be a error probability to SIR that follows a curve from 0 to 100% over a couple of dB variation of the SIR. Also the success rate in this area of the curve will depend on the length of the packet. For" 0*## simplicity reasons this is not included in the model. Instead the normal distribution functions added to both the signal and the interference will more or less serve a similar purpose, and will partly compensate for this. In addition the success rate per packet can be controlled by a "noise" parameter, that allows you to control the percentage of lost packets. This effect will be independent for the actual level of the signal and the packet length. However because the wireless network is expected to be interference bounded instead of noise bounded, this is not considered critical. The same would apply for other effects like channel coherence changes during the transmission of a packet. Although it would not be difficult to more accurately model the described relations, they are considered to have a minor influence on the global accuracy of the results. Model sizing The model allows only up to 2 networks to work in parallel with a limited number of stations per network. Simulation of more than 2 networks is not felt needed, because the 2 network case gives a good feeling for the interference issues involved, and it allows us to determine the boundary between sharing and reuse. Also increasing the number of stations per network, although it could be simply implemented, is not felt relevant. Simulation of network behavior with a large number of stations each contributing a relative low load, is impractical. It is difficult to get meaningful results out of such simulations. It would require very long test time, and it would be difficult to analyze the peak load behavior, because it will only occur temporarily. The proposed method for these large networks is the following: XCharacterize the high load behavior as function of the number of simultaneous active workstations, similar to the tests shown in Doc. IEEE P802.1191/125.(# XAnalyze the traffic pattern of a station needed per transaction/session.(# XUse a Markov chain analyses approach based on this traffic pattern and the total number of stations to determine the probability figures for the concurrent activity of 2 to n stations, and its effect on the performance.(# Example Test reports As already discussed two reports are generated: XAn overview showing multiple test iteration results(# XA detailed report showing results of individual stations, and the counted packet failures by type, and by direction (to or from the server).(# " 0*##Ԍhe appendix shows a sample test report of two networks situated side by side with a large wall or floor in between. The stations 0 and 8 are the Server stations of network 1 and 2 respectively. In addition both networks each have 7 workstations, all generating a high load on the network. Note that in reality this kind of load would typically only occur in probably a few percent of the time in a moderately loaded network with 50100 users. XThe overview report shows the parameters, the topology (only one network is shown), and the throughput (excluding Novell and MAC overhead) of individual networks, as well as the percentage of times that either the packet to the server or the response from the server are in error.(# XThe first detailed report shows a sample with both networks 45 meters apart, behind a wall or on another floor. The columns shown are as follows:(# WS:` `  "The station number(#` TxTot: "Total number of packets transmitted by this node(# Retry:` `  "Total number of times, that the station is forced into backoff Rlim:` `  "Number of times that max. retry limit is exceeded.(# Ccnt:` `  "Collision count from workstation to server JamC:` `  "Jam count from workstation to server(# Lost:` `  "Total number of packets lost from station to server PcK:` `  "Total number of successful packets from this station arrived at the server SPck:` `  "Total number of responses from the server arrived at this station SLst:` `  "Total number of lost packets from server to this station SJmc:` `  "Jam count from server to station SCcnt: "Collision count from server to station TxL:` `  "Transmit level at 1 meter from antenna Del:` `  "Average delay from station to server %Busy "Percentage that medium was busy %Good "Percentage of used bandwidth that was used successfully SRxQ "Receive Queue contents (Instantaneous) STxQ "Transmit Queue contents (Instantaneous) StSLevel "Average Server to server attenuation Plost` `  "Percentage of packets lost Mlost` `  "Percentage of either packet to server or packet from server being lost. A second detailed report shows asimilar report, but now with the Ack protocol enabled. Experience shows that the detailed report is very important,because overall network throughput can still be very good, despite of the fact that individual stations can have low throughputs compared to others. It is also crucial in determining the root cause of the stations behavior, which" 0*## could be resolved by protocol improvements and changes in parameterization. Protocol enhancements One of the protocol enhancements which is included in the model is a MAC level Acknowledge protocol. This can be included by extending the state machine with a few states. Still it is not needed that a separate receiver is modeled. Of course the model also evaluates the possible interference generated by the Ack signal, and the correct reception of the Ack is evaluated in a similar way as done with the packet itself. In addition the detailed report file also gives the number of Ack's that get lost due to interference as is shown in the second detailed report. Note that this means that the actual packet is received successfully, but due to the failure of the Ack, it is still counted as an error, and the packet will be retransmitted. Other protocols Implementation of other protocols requires the implementation of other transmitter state machines. In addition it is possible that more facilities are needed to communicate different conditions between the processes. It is also possible that also receiver processes need to be implemented to allow simulation of all possible effects. For instance the 4 way LBT MSDU format as described in the Ken Biba proposal, needs implementation of receiver functionality. This is because not only the situation at the addressed receiver of the packet must be evaluated as described in this paper. In that protocol at least the RTS and CTS packets cause actions in the other receivers in the network, which is important to control the network access function. This means that for every transmitted packet all receivers need to be activated, and the individual interference conditions need to be evaluated separately. This will make things much more complex and will effect the performance of the simulator. Implementation The simulation program is implemented using the structural Power Basic language. The basic simulation engine uses fixed point computations as much as possible to optimize for speed. Consequently there is not much performance difference when running without a coprocessor. The first operational version was for the CSMA/CA protocol as used within the WAVELAN product. The simulation results were verified against the actual field measurements. Since then more functionality is added regularly. Implementation of other protocols like the 4way LBT protocol is planned. " 0*## Conclusion: A powerfull simulation tool for MAC protocol evaluation in a Radio environment has been constructed. The main characteristics of the PHY have been successfully modeled: XSignal path attenuation as function of distance (# Effect of attenuation boundaries like walls and ceilings XFading / Shadowing(# XCapture effect(# XCochannel interference(# Adjacent channel interference Microwave oven interference (jammer) These characteristics are considered crucial for analyses of the impact of cochannel and adjacent channel mutual network interference. The objective is to evaluate performance aspects of different MAC protocol approaches, and their robustness against interference. The simulator has been designed to analyze the WAVELAN CSMA/CA protocol and is being used to evaluate several protocol alternatives. The model provides simulations at a high traffic load in a realistic ClientServer, and in a PeertoPeer environment. The model allows efficient Analyses of the causes of packet loss at individual stations, by the detailed result report. The model provides the possibility to evaluate the medium reuse aspects, and the relevant protocol characteristics to support it. |0*## #p4  p(ACԼ# #Wireless Network Performance Modeling Approach *Objectives ` ` Develop a performance simulator suitable to evaluate MAC protocol alternatives in an indoor Radio environment.(#` ` ` The simulator should allow us to make the necessary tradeoffs in the development of a efficient and robust wireless protocol.(#` ` ` The relevant PHY characteristics should be modeled to provide a "realistic" environment for performance analyses and parameter tuning.(#` 0*## #Wireless Network Performance Modeling Approach *PHY effects modeled: ` ` Signal path attenuation as function of distance (#` ` ` Effect of attenuation boundaries like walls and ceilings ` ` Fading / Shadowing(#` ` ` Capture effect(#` ` ` Cochannel interference(#` ` ` Adjacent channel interference ` ` Microwave oven interference (jammer) *Other modeling aspects: ` ` Network Topology (location of the Server)(#` ` ` Network Operating System ` ` Type of traffic (R, W, RW)(#` ` ` Peertopeer versus ClientServer traffic ` ` Traffic load ` ` Media Access Protocol (CSMA/CA)(#` J"0*## #Wireless Network Performance Modeling Approach *Model Characteristics: XSingle Network mode:` hh01 Server and up to 15 WS's (workstations)(#h XDual Network mode:` hh02 Networks with 1 Server and up to 7 WS's(#h XAllows entry of station location coordinates for one Network.(# XSecond network has the same topology but the distance between the two networks can be varied in any direction.(# XExtra attenuation offset between the two networks to simulate effect of a wall or operation on a different floor.(# Two Traffic Model versions are supported: .` ` "Peer to peer" traffic heavy load performance(#` .` ` Client Server traffic heavy load performance(#` Three different traffic modes for the Client Server model: .` ` Continuous Read .` ` Continuous Write .` ` Random Read / Write XCarrier Sensing based on Path attenuation between any station.(# J"0*## #Wireless Network Performance Modeling Approach XCapturing effect Included.(# .` ` Both packets lost when separation receiver < SIR (#` .` ` One packet lost when separation > SIR, while the other packet separation < SIR.(#` .` ` No packets lost when destination addresses are different and both meet separation > SIR condition.(#` XCochannel Interference model based on Path attenuation and SIR separation requirement.(# XA Normal Distribution "Fading Margin" uncertainty applied for all Path attenuations calculations. (# XProvide means to induce lost packets.(# XA Micro Wave oven interferer (jammer) with programmable onoff duty cycle at a programmable level.(# XAn Adjacent Channel environment isolation between the channels controlled by a parameter. (No carrier sensing)(# XTraffic load controlled by fixed + random delay per station.(# XAll CSMA/CA parameters and PHY parameters like transmit level and Carrier Sense level can be controlled.(# X"Novell retry timeout" applied.(# 6 0*## #Wireless Network Performance Modeling Approach XMonitors packet lost tallies per link.(# Average MSDU delay calculated per individual station. XPerformance measured in KBytes/sec actual data throughput excluding Novell overhead in the Perform3 test.(# XPeer to peer performance measured including overhead, and without NCP handshaking. Traffic destinations are random. (# XHigh performance "Event driven" Simulator.(# XProduces two different output files:(# .` ` Detailed report showing performance and lost packet statistics of individual Stations separated in "To the Server" and "From the Server" directions in each network.(#` .` ` Summary report showing throughput and Collision probability per network as a function of an iteration parameter like "Distance between Networks".(#` 0*##  r5 h   #d6X@Kț@#AppendixA ` `  PERForm3 Test (Client Server) Date: 02-26-1992 Time:10:41:46 File: Perfr63 .dat ******************************************************************* ************ Two networks separated by wall/floor ******************************************************************* Performance test 2 Networks with 7 Workstations each Traffic Mode = R Attenuation Coefficient = 3.5 Fading Margin = 5 dB Normal Distributed Extra noise outage = 0 % SIR = 10 dB Carrier Sense level = -82 dBm Server delay = .5 msec + random .5 msec Workstation delay = 1 msec + random .5 msec Packet Data length = 512 Bytes Lost Packet Timeout = 300 msec Test time = 5 sec Attenuation offset between Networks = 20 dB ******************************************************************* . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 . . . 6 . . . . . . . . . . . . . . . . . . . . . 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 . . . . . . . . . . . . . . . . . . . . . . . . . 2 . . . . 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Scale = 10 meters per dot. ******************************************************************* Network-1` `  ")hh0Network-2 Separation Throughput %Lost Throughput %Lost Total 5.0 77.3 10.2 85.5 9.2 162.8 R 25.0 126.1 3.0 124.6 3.3 250.7 R 45.0 128.4 4.8 135.4 3.6 263.8 R 65.0 141.0 2.3 137.0 3.3 278.0 R 85.0 138.1 2.8 138.0 2.6 276.2 R f0*%% ` `  " PERForm3 Test (Client Server) Date: 02-26-1992 Time:10:39:51 File: Perf63.dat ******************************************************************* ************ Two networks separated by wall/floor ******************************************************************* Performance test 2 Networks with 7 Workstations each Traffic Mode = R Attenuation Coefficient = 3.5 Fading Margin = 5 dB Normal distributed Extra noise outage = 0 % SIR = 10 dB Carrier Sense level = -82 dBm Server delay = .5 msec + random .5 msec Workstation delay = 1 msec + random .5 msec Packet Data length = 512 Bytes Lost Packet Timeout = 300 msec Distance between Networks = 45 0 meter Attenuation offset between Networks = 20 dB ******************************************************************* WS TxTot Retry Rlim Ccnt JamC Lost Pck SPck SLst SJmc SCcnt TxL Del R 0 1278 468 0 0 0 0 0| 0 0 0 0 -6 1 254 210 0 4 1 5 249| 246 3 0 3 -6 2.56 2 182 170 0 9 0 9 173| 172 0 0 0 -6 2.86 3 198 183 0 7 0 7 191| 188 2 0 2 -6 2.81 4 233 342 0 3 0 3 229| 225 4 0 4 -6 4.64 5 104 202 0 5 0 5 99| 93 6 5 1 -6 5.63 6 86 196 3 4 0 7 79| 73 6 5 1 -6 4.90 7 265 262 0 5 0 5 259| 257 2 0 2 -6 3.03 Totals 3 37 1 41 1279 1254 23 10 13 0 2.63 TimeStamp %Busy %Good SRx-Q STx-Q StS-Level Throughput 128.4 KByte 500000 73.4 99.5 0 1 -88 PLost 2.5 % MLost 4.8 % WS TxTot Retry Rlim Ccnt JamC Lost Pck SPck SLst SJmc SCcnt TxL Del R 8 1338 473 0 0 0 0 0| 0 0 0 0 -6 9 86 295 5 2 0 7 78| 74 4 4 0 -6 8.64 10 170 234 0 4 0 4 166| 162 4 0 4 -6 4.18 11 238 208 0 4 0 4 234| 231 2 0 2 -6 2.56 12 213 239 0 6 0 6 207| 206 0 0 0 -6 3.40 13 171 199 1 7 0 8 163| 162 1 0 1 -6 3.02 14 195 208 0 3 0 3 191| 187 4 0 4 -6 3.22 15 304 321 0 3 0 3 300| 300 0 0 0 -6 3.07 Totals 6 29 0 35 1339 1322 15 4 11 0 2.63 TimeStamp %Busy %Good SRx-Q STx-Q StS-Level Throughput 135.4 KByte 500000 76.8 99.6 0 0 -88 PLost 1.8 % MLost 3.6 %0*%% ` `  "PERForm3 Test (Client Server) Date: 02-26-1992 Time:10:54:48 File: Perf63.dat ******************************************************************* ************ Two networks separated by wall/floor ************ With Ack Protocol Enabled and max retry limit= 20 ******************************************************************* Traffic Mode = R Attenuation Coefficient = 3.5 Fading Margin = 5 dB Normal distributed Extra noise outage = 0 % SIR = 10 dB Carrier Sense level = -82 dBm Server delay = .5 msec + random .5 msec Workstation delay = 1 msec + random .5 msec Packet Data length = 512 Bytes Lost Packet Timeout = 300 msec Distance between Networks = 45 0 meter Attenuation offset between Networks = 20 dB ******************************************************************* WS TxTot Retry Rlim Ccnt JamC Lost Pck SPck SLst SJmc SCcnt AckL TxL Del R 0 1332 576 0 0 0 0 0| 0 0 0 0 0 -6 1 200 291 1 6 0 7 199| 198 4 0 4 0 -6 3.89 2 186 310 1 9 0 10 185| 184 2 0 2 0 -6 4.72 3 212 316 0 10 0 10 212| 211 5 0 5 0 -6 4.33 4 183 369 1 5 0 6 182| 181 3 0 3 0 -6 5.76 5 194 396 0 10 0 10 194| 193 8 2 6 0 -6 6.50 6 153 381 3 6 0 9 149| 149 7 6 1 0 -6 6.73 7 216 290 0 1 0 1 216| 215 1 0 1 0 -6 4.03 Totals 6 47 0 53 1337 1331 30 8 22 0 2.80 TimeStamp %Busy %Good SRx-Q STx-Q StS-Level Throughput 136.3 KByte 500000 83.9 99.4 0 5 -93 PLost 3.0 % MLost 6.0 % WS TxTot Retry Rlim Ccnt JamC Lost Pck SPck SLst SJmc SCcnt AckL TxL Del R 8 1346 559 0 0 0 0 0| 0 0 0 0 0 -6 9 54 470 8 1 0 9 45| 45 1 0 1 0 -6 23.89 10 198 337 1 5 0 6 196| 196 3 2 1 0 -6 4.88 11 231 312 0 7 0 7 230| 230 2 0 2 0 -6 3.84 12 224 314 0 5 0 5 224| 223 3 0 3 0 -6 4.16 13 220 328 0 3 0 3 220| 219 4 0 4 0 -6 4.49 14 210 298 1 5 0 6 208| 208 3 0 3 0 -6 3.94 15 226 300 0 4 0 4 226| 225 1 0 1 0 -6 4.00 Totals 10 30 0 40 1349 1346 17 2 15 0 2.77 TimeStamp %Busy %Good SRx-Q STx-Q StS-Level Throughput 137.8 KByte 500000 84.1 99.6 0 3 -93 PLost 2.1 % MLost 4.1 %