FRACAS FRAmed Channel Access Simulator USER MANUAL
Contents
1. line The header and the blank line are not printed with the t terse option syntax computer lt stations gt lt observable gt listing ofile lafile lopipe type every where type a string whose value is normal aggregate or compact See above for explanation Default normal every a positive integer See above for explanation Default 1 5 7 2 4 Distribution computer categorise This computer divides the samples of the observables into categories or bins The number of categories is a parameter All the categories have the same size and span the entire range of the observable from its minimum to its maximum value When zero is specified as the number catn of categories the categories are built with a width equal to 1 centred around all integer values plus 0 5 covering the entire range of the observable This feature is useful when the observable is known to take on integer values The maximum number of bins computed when catn is 0 is max_cslen a global variable to be specified at the beginning of the input file If the range of the observable is greater than this parameter s value the bins are enlarged to integer widths using the smaller width such that the number of categories is less than max_cslen A list of lines is produced with two values in each line the centre value of the category and the number of samples in it The sum of all the samples in the different categories is equal to t2. A datagram request in a frame can only be made if the station has some allocation either stream VBR or datagram in that frame If not no datagram request is made 3 3 feeders drifs FEEDERS and DRIFS protocols This requester is supposed to be used together with the feeders or the drifs allocators see 4 3 and 4 4 When used together they behave like the FEEDERS or DRIFS protocols respectively The stream VBR and datagram requests are made in the same way as the fodaibea requester does The only difference is that in every frame a datagram request is always made 4 ALLOCATION ALGORITHMS An allocator is a function called at each frame It looks for the requests that the stations made a given number of frames before the number depends on the allocator type and computes the allocations for all the stations Only one allocator can be specified 4 1 fixed Fixed TDMA The stream traffic receives an assignment equal to the request sreq The VBR traffic receives an allocation equal to the max VBR request vmaxreg The remaining space equal to framesize minus the sum of the stream and VBR allocations is evenly divided among all the stations and assigned as a datagram allocation 4 2 fodaibea FODA IBEA and FID VBR protocols FODA IBEA is a satellite access scheme for simultaneous transmissions of both stream and datagram data The quality of service is maintained even when the system is faded i e the transmission
3. THE INPUT FILE The input file consists of a sequence of one or more run descriptions The run descriptions are sequenced and separated by a line containing only the end keyword Putting several run descriptions into a single file is the same as putting them into separate files and writing the file names on the FRACAS command line A comment can be put anywhere in the input file All characters in a line which follow a sign are ignored Also entire lines whose first character is are ignored i e those lines are not considered as blank lines This difference is important because blank lines are significant in the input file syntax 5 1 GLOBAL VARIABLES One global variable per row must be specified in the input file framesize Size of the frame expressed in TRUs frametime Time length of the frame expressed in TMUs rttime Round trip time expressed in TMUs histlen Size of the memory allocated to record the histories expressed as a number of samples A sample per frame is generated for each observable statistic different from tru_delay To be removed in a future version trudhistlen Size of the memory allocated to store the history of tru_delay expressed in number of samples A sample is generated for each group of consecutive TRUs sharing the same delay Default 2 histlen warmup Number of TMUs to expire before starting to record the data Default 0 seed Seed for all the random number generations tha
4. round trip time plus one frame No reference burst is transmitted A control slot is permanently assigned to each active station in a position in the allocation cycle that is fixed provided that no station enters or leaves the system The control slot is used to send the requests Only stinframe slots are accommodated in each frame 8 by default Therefore the allocation cycle length is given by the number of active stations divided by stinframe rounded up A fixed quote dquote of the request is given to each station If the sum of the assignments is greater 10 than the length of the allocation cycle then the assignments are compressed to fit into the assignment cycle space otherwise the excess space is divided among all the stations No assignment is less than mindall equal to the burst overhead length bovh by default and maxdall is the maximum allocation equal to half an allocation cycle by default 4 4 FEEDERS This too is a distributed control assignment algorithm 6 A reference burst is transmitted every na frames by a master station na is also the length of the allocation cycle Every station receives an allocation in every frame The allocations are changed only once every na frames There is no minimum allocation but a maximum allocation maxdall equal to half a frame length by default A mechanism similar to the one used by DRIFS is used to compute the allocations using a fixed quote dquote of the request 11 5
5. signal is attenuated due to bad atmospheric conditions VBR traffic is not supported 1 2 3 FID VBR is a version of FODA IBEA modified in a such a way as to support VBR traffic too 4 5 Both access schemes FODA IBEA and FID VBR have centralised control In terms of delay the delay between a request and the relevant assignment is twice the round trip time rttitme divided by frametime rounded up plus three This accounts for all the transmission and the processing delays The channel overhead is due to a traffic independent part consisting of the reference burst an FAS First Access Slot every 32 frames by default and some extra space needed by the hardware The reference burst is sent by the control station for synchronisation and contains the assignments The FAS is a space used by the stations to enter the satellite network Other overheads are traffic dependent the only one FRACAS considers is the burst overhead one per frame which is the unused space subtracted from the total frame space every time an allocation is given to a station The fodaibea allocator implements FID VBR as a superset of FODA IBEA However since FRACAS has no means to change the signal quality support for variable satellite channel speed and coding rate is not implemented The stream allocation is equal to the stream request The VBR allocation is equal to the VBR request The datagram requests are organised into a ring which is scanned to co
6. space to datagram traffic Here is how the allocation sequence for each station works 1 The stream TRUs are sent If s queue is less than s allocation v_extraspace for this frame is set to the difference and s_sent is set equal to s_queue Otherwise s_sent is set equal to s_allocation 23 2 The VBR TRUs are sent If v gueue is less than v_allocation plus v_extraspace d_extraspace for this frame is set to the difference and v_sent is set equal to v_queue Otherwise v_sent is set equal to v_allocation plus v_extraspace 3 The datagram TRUs are sent If d_queue is less than d_allocation plus d_extraspace d_sent is set equal to d_queue Otherwise d_sent is set equal to d_allocation plus d_extraspace For the pseudo station sum the observable is the sum of the extra spaces of all the stations d_unused The space in TRUs unused in a frame by a station It is equal to d_allocation plus d_extraspace minus d_sent For the pseudo station sum the observable is the sum of all the TRUs not used in the system s_delay v_delay d_delay The frame delay in TMUs for each type of traffic Each sample is the mean of the delays of the TRUs generated in a frame by a station s generators The delay is equal to the round trip time plus the queueing time which is an integer number multiplied by the frametime For the pseudo station sum the observable is the mean delay of the TRUs generated in a frame by all the stations s_trudelay
7. stop parameters are the same as in the constant traffic generator Seed is an integer All the other parameters are floating point values syntax generator traffic type poisson traffic ltfactor seed burst start stop cycle where start see explanation for the constant traffic generator Default 0 stop see explanation for the constant traffic generator Default 0 which means never stops cycle see explanation for the constant traffic generator Default 0 which means infinite cycle 5 6 5 3 Impulsive traffic generator impulse This is a two states Markov modulated Poisson generator Generates TRUs with exponentially distributed interarrival times where the exponential constant depends on the state where the traffic generator is The probability of switching to the other state is constant so the time of permanence in each state is exponentially distributed The resulting traffic is Poisson distributed with two possible transmission rates chosen at random Four parameters are needed to define the statistical behaviour of such a generator One is the average generated traffic which is specified using the trafficltfactor parameter The other three parameters are the mean cycle length cycle which is the sum of the mean times of permanence in the two possible states the mean duty cycle duty which is the ratio between the mean permanence in state 1 and the mean cycle length the burstiness burs
8. than framesize and not greater than CS size in TRUs of the First Access Slot Must be less than framesize Default 0 number of frames lasting between two consecutive FASs Default 32 minimum datagram allocation in TRUs Default lt bovh gt maximum datagram allocation in TRUs Must be greater than lt bovh gt Default is framesize 2 minimum number of stations used for computing the quote of the datagram request that is assigned Default value 5 maximum number of stations used for computing the quote of the datagram request that is assigned Default value 50 the quote of the datagram request assigned to each station in the assignment cycle is equal to 100 times the number of stations divided by alld default 100 The minimum and maximum number of stations used for these calculations are allnmin and allnmax respectively defaults 5 and 50 flag for enabling tracing y n Default n 15 16 5 4 3 FEEDERS feeders The channel space is shares among the stations according to the FEEDERS access scheme The stream allocations are equal to the sreg value of each station the VBR allocations are equal to the VBR request bounded by the vminreg and vmaxreg values of the vbrreq keyword the datagram allocations are computed by using the FEEDERS algorithm The delay between the request and the corresponding allocation is an integer number of frames equal to 2 rttime frametime where the division is an intege
9. the datagram assignment cycle bovh burst transmission overhead in TRUs Must be less than framesize maxdall maximum datagram allocation in TRUs Must be greater than lt bovh gt Default is half the available space trace flag for enabling tracing y n Default n 5 5 STOPPER LINE This line specifies the name of a routine that is called at each frame Each time it is invoked the stopper decides whether or not the simulation must end 5 5 1 Elapsed time stopper stopper Interrupts the simulation after a specified time interval syntax stopper maxtime time frames where time number of TMUs after which the simulation must be stopped frames number of frames after which the simulation must be stopped 5 6 STATION DESCRIPTION 5 6 1 The station identifier station syntax station number number Where lt number gt station number If n n is specified the current station block definition is valid for stations from n up to n2 The stations must be declared in ascending order starting from 1 No holes are allowed in the numbering 5 6 2 The stream allocation request streamreq For each station a request for stream traffic allocation can be specified Allocators and Requesters use this number for their computations if they need them syntax streamreq sreq where sreq number of TRUs requested for the stream traffic in each frame 17 5 6 3 The VBR allocation requ
10. v_trudelay d_trudelay The TRU delay in TMUs for each type of traffic This observable provides a sample per TRU while the frame delay observable provides one sample per frame The delay is equal to the round trip time plus the queuing time which is an integer number multiplied by the frametime For the pseudo station sum the observable is the delay of each packet generated in the system The ordering does not make much sense it is the listing of the delay of packets generated by each station in turn so making a graph or correlation of it would yield nonsense 5 7 2 The computers computer is a routine that performs calculations on some station s observables A t tine that perfi lculat tation s ob bl usually some kind of statistics are computed A computer line can specify more than one station The lt stations gt argument can be either a number the station number or two numbers separated by columns the first and the last station numbers In this case the computer is invoked repeatedly on each of the stations specified As a special case the pseudo station sum can be specified Each observable has a special meaning for this pseudo station which is described in the section about observables One more special case all exists the computer is run on all the stations including the sum pseudo station The pseudo station sum is indicated as station number 0 in the output listings All computers produce their output on the stand
11. value of each station The VBR allocations are set equal to the VBR request bounded by the vminreg and vmaxreq values of the vbrreq keyword The remaining space in the frame is equally divided among the stations and assigned as a datagram allocation independently of the stations datagram requests syntax allocator fixed 5 4 2 FID VBR fodaibea The channel space is shared among the stations according to the FID VBR access scheme The stream allocations are set equal to the sreg value of each station The VBR allocations are set equal to the VBR request bounded by the vminreg and vmaxreg values of the vbrreg keyword The datagram allocations are computed 14 by using the FID VBR algorithm The delay between the request and the corresponding allocation being assigned to the requesting station is an integer number of frames equal to 3 2 rttime frametime where the division is an integer excess division syntax allocator fodaibea bovh cs csovh csevery fas fas_every mindall maxdall allnmin allnmax alld trace where bovh cs csovh fas fas_every mindall maxdall allnmin allnmax alld trace overhead in TRUs for burst transmission Must be less than framesize size in TRUs of the control slot Must be less than framesize overhead in TRUs for all the allocations given to a station as redistribution gift or control slot Must be less
12. 3 2 FID VBR fodaibea A station makes a request only when it has an allocation either stream VBR or datagram For the stream traffic a request is made equal to the sreg parameter of the streamreq keyword For the VBR traffic a request is made equal to the 13 vminreq parameter of the vbrreg keyword if the input traffic is less than vminreg otherwise the request is equal to vmaxreg The VBR input traffic estimation is changed at each frame using the average value of the last vwin frames The datagram request is equal to the datagram queue length plus dH times the input traffic where dH is a temporal constant The datagram input traffic estimation is changed every dwin frames using the average value on this time interval syntax requester fodaibea dH dwin vwin where dH temporal constant Non negative floating point number dwin positive number Default 1 vwin positive number It may be omitted if no station specifies vminreg and vmaxreq 5 3 3 FEEDERS and DRIFS Same as the fodaibea requester The only difference is that a request is always made irrespective of whether the station has an allocation or not 5 4 ALLOCATOR LINE Only one allocator can be specified It computes the allocations of each station It is called once per frame it looks at the requests of all the stations and computes the allocations of all the stations 5 4 1 Fixed TDMA fixed The stream allocations are set equal to the sreg
13. 5 6 5 4 Fractional Gaussian Noise traffic generator ESTP ct EA OE AN EE AA EE OO ts state 21 5 6 5 5 External traffic generator external 21 3 1 COMPUTER LINES stet kes ve ae Pa eg ees ee ok eg Ge NA KRAG ee ee 22 SA ES H TE T DED EE E uko Ku A EA 22 31 2 ThE COMPULETS pini 20493 A ee KARG ee ee Re AIAR ken aa DE eo 24 5 7 2 1 Simple statistics computer simplestats 25 5 7 2 2 Quantile computer quantile n 25 5 7 2 3 Sample listing computer listing 26 5 7 2 4 Distribution computer categorise 26 5 7 2 5 Computer for graphing the distribution on a tty ti GISEIAD oe Ne De GR e i i no 27 5 7 2 6 Computer for graphing the observable versus time on a tty ttygraph esse ee see ee ee ee Re ee ee ee ee 27 5 82 INPUT FILE SYNTAX a aa 28 REFERENCES N EE RE OR RE EE RE atten 30 du d ole B SE EE Lr 31 An example input Hie is RS ARE ER RR RE ND AG MIGRA DASA RARAS bek ee NARRATA 31 An example FRACAS output iese sesse ese ee ee ee ee ee Ge ee ee 32 iii LEGEND a is optional one or more instances of a alpl lf one and only one item among a b f TRU Traffic Unit Data size unit measurement TMU time Unit Time unit measurement The traffic unit measurement is TRU TMU observable any measurement on Which statistics can be collected history the v
14. FRACAS FRAmed Channel Access Simulator USER MANUAL Nedo Celandroni Erina Ferro Francesco Potorti CNUCE C N R Institute Via S Maria 36 56126 Pisa Italy Tel 39 50 593207 312 203 Telex 500371 CNUCE Fax 39 0 50 904051 052 e mail n celandroni e ferro pot cnuce cnr it CNUCE Report C95 27 November 1995 TABLE OF CONTENTS TABLE OF CONTENT S iL ER RIGA ARIAS ees EE ee i INTRODUCTION set ee GR Ge De ARAS KEK Ge Se KLAREGA DAMA A PAR be As De ge ek bek AAJ 1 2 OVERVIEW OF PRACAS ses N Ek EG ERARA ERA KERA EK RNA KANA see ee 3 3 REQUEST ALGORITHMS iese ase es Ee Me RA AG RARA RAGA ARIA ee be ke ke RE Ne ee 7 Se Li RE EE EE EE EE EE EE EE EE ORR RE OE EER EE EE EEN 7 3 2 fodaibea FODA IBEA and FID VBR protocols 7 3 3 feeders drifs FEEDERS and DRIFS protocols ee see eg 8 4 ALLOCATION ALGORITHMS eee esse esse ese ese ee ee ee ee ee ee Ge ee ee Ge 9 4 1 fixed Fixed TDMA ees ee Re ee Re ee Re ee ee 9 4 2 fodaibea FODA IBEA and FID VBR protocols 9 da B dl RE EO EE NALI 10 A AL RREDERS AR AA AR EE EE ER N 11 o THE INPUT EILE ER EI AGA KASA Ek AI Ges AR Ee ES RE AR GEE Gee ee 12 5 1 GLOBAL VARIABLES sesse be ee ees bees kas ke EER oe Sea krab kas kena kaa pe a o 12 52 NITER EN E ee eN PPT ese Ra ee N Ne es AR Ge See ee ee ene 13 5 2 1 Equally distributed
15. alues assumed by an observable 1 INTRODUCTION FRACAS FRAmed Channel Access Simulator is a simulation tool for TDMA satellite networks which provides users with a library of satellite channel access schemes in TDMA Once the network configuration has been defined the user can choose the band allocation policy from a set of predefined policies Statistics on the performance of the chosen access scheme in the conditions of the network simulated are then collected FRACAS is aimed at all those research centres in the field of communications via satellite in TDMA and to service suppliers who need support from simulative studies in order to choose the best allocation policy and to tune up the relevant parameters FRACAS is useful for comparing different satellite access schemes in traffic load conditions chosen by the user Three classes of traffic called datagram stream and VBR are considered Datagram includes all the jitter tolerant applications stream and VBR include all the real time applications Datagram traffic is a connectionless type of traffic without any particular delay requirements It can tolerate out of order delivery of packets and a high jitter but usually it requires a low bit error rate The delay introduced by the network s crossing is not a critical constraint However especially on a high delay network like the satellite network the end to end throughput of such traffic can be heavily impaired by bit errors
16. ard output by default Each computer can also be instructed to send its output to a file or to an external program One of three arguments can optionally be added to the argument list of each computer ofile send the output to a file specified as an argument The previous contents of the file are overwritten afile append the output to a file specified as an argument opipe pipe the output to a program whose command line is specified as an argument For each of the three possibilities a string must be supplied after the equal sign The string may be enclosed between double quotes if it contains other than alphanumeric characters plus the period and the minus sign A double quote within the string can be quoted by preceding it with a backslash The output produced on the standard output is generally more verbose than the output directed to files or external programs Details are provided in each computer s description It is also possible to obtain a terse printing on standard output by invoking FRACAS with the r terse option In this case the output sent to stdout is equal to the output sent to files or pipes 5 7 2 1 Simple statistics computer simplestats This computer prints on standard output the minimum maximum number of samples mean and standard deviation of the specified observable These statistics are printed on a single line preceded by a header line and followed by a blank line The header and the blank line are no
17. datagram allocation initer even 13 5 2 2 No datagram allocation initer Zero ee RR 13 33 REQUES TER LINE 5e kes Ve es ese ee ees Ne ke ERAO ee be ee eg RA ASI 13 5 3 1 Queue requester QUE E ses eN bees Ee ii 13 5 3 2 FID VBR iodaibER isnie eg bek bes AE ee Re ee IA aa 13 9 39 FEEDERS and DRIBS asse Gees TA E es LERA ARAS ee Ges ee Re 14 SAAL OGATOR LINE i ee ee ee ee n IAA Ne Ek ee ee es ee ng ee 14 5 4 1 Fixed TDMA fixed iii 14 5 4 2 FID VBR fodaibea i 14 5 4 3 FEEDERS feeders ii 16 34 4 DRIFS EIS ees es es ana 16 59 STOPPER LINE pila ee AE ee RARAS at 17 5 5 1 Elapsed time stopper Stopper se ee Re AR Re 17 5 6 STATION DESCRIPTION ies soes ees es soe ge es Re RA eg ee Re eg Pe Re ge 17 5 6 1 The station identifier Station i 17 5 6 2 The stream allocation request streamreg ee il 17 5 6 3 The VBR allocation request vbrreg n 18 5 6 4 The maximum queue length se ee ee ee ee ee ee 18 5 6 5 The traffic generators i s es ne Re AR ese be aaa 18 5 6 5 1 Constant traffic generator constant 19 5 6 5 2 Poisson traffic generator POISSON ees sesse ee iic il 20 5 6 5 3 Impulsive traffic generator impulse 20
18. e global variable seed In this case the seed is generally different if the overall structure of the simulation changes that is if stations are added or removed or the generators changed or anything else changes Moreover if the global seed parameter is 0 and the generator seed parameter is 0 the seed used for the pseudo random number generation changes at each run provided that the runs are at least one second apart because the global seed variable is initialized with the system time If the same sequence of traffic from a generator in different runs is required a seed different from must be set The default for the seed parameter is 0 Each generator generates bursts of TRUs In the generator descriptions whenever a TRU is mentioned a burst of TRU is meant if the burst parameter is set to a number greater than 1 The rate of the burst generation remains the same whatever the burst length is The number of TRUs generated in a frame is simply multiplied by the burst size an integer number The default burst length is 1 Each generator has a parameter that indicates the mean total traffic generated TRUs per TMU This can be either a traffic floating point number indicating the absolute value of the traffic mean or a factor floating point number indicating the traffic mean relative to the global parameter ref_traffic In the first case the traffic generated is equal to traffic burst in the second case it is equal to actor re
19. ed by assigning peak rate bandwidth to VBR video applications and by using the residual bandwidth for non real time traffic 2 OVERVIEW OF FRACAS FRACAS is a batch program that reads its input from one or more files and outputs the results onto one or more files or terminals The general syntax for invoking FRACAS is summarised in the following screenful which FRACAS prints when it is called with no arguments usage fracas options files Where options are e print all known values on stdout skip all computations s print the entire source file before the computation results t terse do not print headers on standard output y verbos write messages showing the process of computations At least one input file name must be specified on the command lines but several are accepted The input files are individually read and elaborated by FRACAS The e emulate only option skips all the computations on the output values runs the emulator and writes all observable values on the standard output Only the packet delay observable tru_delay is not printed because it requires a computation to be evaluated The e option is useful for using FRACAS as an emulator and for giving the task of computation to some other program The s source file option is useful when the output of FRACAS is redirected to a file that at the end of the run will contain both the source of the run and its results The t terse option
20. eloped in the framework of the Olympus project and its validity has been confirmed by simulative and experimental results FODA IBEA does not support VBR traffic FID VBR is the FODA IBEA protocol with support for VBR video traffic DRIFS and FEEDERS are distributed control protocols Work on FRACAS was begun to evaluate the performance of these last two protocols and it is expected to grow to include other protocols which can be easily added by coding them in C Computers Once the allocation policy has been chosen various statistics such as minimum maximum average variance percentile and sample listing on the following observables can be collected input traffic Jength of the input queues output traffic unused allocation space frame by frame and packet by packet delay packets lost due to internal buffer overflow Additional statistics on the observables listed above can be added by coding the relevant routines in C The end keyword The end keyword on a line after the computers section means that the description of a run is finished and another one follows Therefore an arbitrary number of runs can be concatenated into a single file The last file needs no end keyword To concatenate runs into a single file is the same as writing them in different files and listing the file names on the FRACAS command line 3 REQUEST ALGORITHMS A requester function is called at each frame for each station It computes the r
21. equest the station makes in that frame for each of the three kinds of traffic stream VBR datagram All the stations in FRACAS share the same allocator so only one allocator needs to be specified 3 1 queue All the requests are equal to the length of their respective input queues This requester is only useful for simple tests on the behaviour of the FRACAS simulator itself 3 2 fodaibea FODA IBEA and FID VBR protocols This requester is supposed to be used together with the fodaibea allocator see below When used together they behave like the FODA IBEA or FID VBR protocols The stream request is equal to sreg as specified in the station definition The VBR request is equal to the VBR minreq or maxreq specified in the station definition The maxreq value is requested when the input traffic is greater than the VBR minreg value The input traffic is measured as the average input in a sliding window whose length in frames is the fodaibea requester parameter vwin The minreq value is requested when the input traffic is less than or equal to minreg The datagram request is proportional to the traffic coming into the station plus the backlog i e the volume of data waiting for transmission to the satellite request H traffic backlog where H is a temporal constant The input traffic is updated every dwin frames it is computed as the average of the input in the last dwin frames Both H and dwin are parameters of the requester
22. erators and adds them to its input queues one for each type of traffic The lengths of these queues are registered in the observables above the queues are then reduced by the number of TRUs to be sent in the current frame For the pseudo station sum the observable is the sum of the queues of all the stations s_sent v_sent d_sent Number of TRUs transmitted in a frame by a station If the length of the queue of a specific type of traffic is greater than the relevant allocation received this number is set equal to the allocation for that frame otherwise it is equal to the queue length for that frame For the pseudo station sum the observable is the sum of the TRUs sent by all the stations s_request v_request d_request Request of a station in a frame for each type of traffic These are the numbers computed by the requester at each frame For the pseudo station sum the observable is the sum of the requests made by all the stations s_allocation v_allocation d_allocation Number of TRUs allocated to a station in each frame for each type of traffic These are the numbers computed by the allocator For the pseudo station sum the observable is the sum of the allocations given to all the stations v_extraspace d_extraspace There is a hierarchy among the types of traffic At each frame the space not used by stream traffic is given as extra space to the VBR traffic and the space not used by VBR traffic is given as extra
23. est vbrreq For each station two requests for VBR traffic allocation can be specified Allocators and Requesters use these numbers for their computations if they need it syntax vbrreq vminreq vmaxreq where vminreq is the minimum number of TRUs requested in each frame vmaxreq is the maximum number of TRUs requested in each frame 5 6 4 The maximum queue length For each station the maximum lengths in TRUs of the stream VBR and datagram queues can be specified Zero means infinite queue When a queue length is different from 0 incoming TRUs are rejected once the respective queue length has exceeded its maximum length The number of packets discarded is registered in the s_dropped v_dropped and d_dropped observables respectively syntax maxqueuelen s v d where s maximum stream input queue length Default 0 v maximum VBR input queue length Default 0 d maximum datagram input queue length Default 0 5 6 5 The traffic generators For each station several traffic generators can be defined The available generators are constant poisson impulse fgn external Each generator generates one traffic type chosen from datagram d stream s or VBR v Each non deterministic generator has a seed parameter which can optionally be set to an integer The seed is used to generate pseudo random numbers If a seed of 0 is given the seed for that generator is generated at startup by using th
24. f_traffic burst 5 6 5 1 Constant traffic generator constant This generates TRUs with constant interarrival times The use of the start stop and cycle optional parameters allows an On Off generator to be created with a constant duty cycle and fixed interarrival times between TRUs during the On states cycle is the repetition period length A cycle is composed of an Off On off sequence Within the cycle start and stop are the times when the On period begins and ends respectively cycle set to 0 means no repetition infinite period length Stop set to 0 means that the on period lasts until the end of the cycle Thus an infinite step of constant traffic has both cycle and stop equal to 0 while a single impulse of constant traffic has cycle equal to 0 and stop different from 0 All the parameters are floating point values syntax generator traffic type constant traffic ltfactor burst start stop cycle where start see explanation above Default 0 stop see explanation above Default 0 which means never stops cycle see explanation above Default 0 which means infinite cycle 19 5 6 5 2 Poisson traffic generator poisson This generates TRUs with exponentially distributed interarrival times Start stop and cycle optional parameters allow an On Off generator to be created with a constant duty cycle and a Poisson generation rate during On states The meanings of the cycle start and
25. he number of samples of the observable The listing is preceded by a header line and followed by a blank line The header and the blank line are not printed with the t terse option syntax computer lt stations gt lt observable gt categorise ofile lafile lopipe catn where catn number of categories in which the range of observable values is divided 0 makes categories of width 1 centred on integer numbers plus 0 5 Default is 0 5 7 2 5 Computer for graphing the distribution on a tty tfydistrib This computer creates a graph printable on a tty or a character printer which represents the mass density function of the observable The graph is preceded by a header line and followed by a blank line The header and the blank lines are not printed with the t terse option syntax computer lt stations gt lt observable gt ttydistrib ofile lafile lopipe rows cols where rows number of rows used for representing the character graph Default 22 cols number of columns used for representing the character graph Default 78 5 7 2 6 Computer for graphing the observable versus time on a tty ttygraph This computer creates a graph printable on a tty or a character printer which represents the observable versus the time The graph is preceded by a header line and followed by a blank line The header and the blank lines are not printed with the t terse option syntax computer stations l
26. ics to be collected The syntax is described in the Appendix The main entries in the input file define the time length and the data size of the frame the configuration of each station the channel allocation policy the statistics to collect the duration of the simulation the end keyword e Station Configuration For each station the following parameters can be defined maximum queue length for stream VBR and datagram stream request maximum and minimum VBR request For each station an arbitrary number of different traffic generators can be defined Examples of traffic generators are Constant rate Periodic on off with constant rate bursts Poisson Two state Markov modulated Poisson Fractional Gaussian Noise Additional traffic generators can be added by coding them in C language Channel Allocation Policy Users can choose the channel allocation policy for a simulation session The available policies at the time of writing include Fixed TDMA FODA IBEA Fifo Ordered Demand Assignment Information Bit Energy Adapter FID VBR FODA IBEA Derived Variable Bit Rate DRIFS Distributed allocation with Requests In Fixed Slots FEEDERS Faded Environment Effective Distributed Engineering Redundant Signalling Apart from Fixed TDMA which is a well known policy the others were studied at CNUCE In particular FODA IBEA which has a centralised control was dev
27. ironment CNUCE Report C 95 28 September 1995 An example input file point 1 6 framesize 615 frametime 20 rttime 252 histlen 300000 warmup 12000 seed 0 initer even ref_traffic 0 6 APPENDIX traffic units frame length is 20ms round trip time is 252ms length of histories wait 12s before registering requester feeders drifs dH 400 dwin 5 vwin 25 allocator drifs dquote 0 05 bovh 8 stinframe 8 stovh 16 stopper maxtime time 6000000 station 1 generator d impulse station 2 generator d impulse station 3 generator d impulse cycle 3000 duty 0 15 burstiness 5 burst 4 tfactor 0 42 cycle 3000 duty 0 15 burstiness 5 burst 4 tfactor 0 31 cycle 3000 duty 0 15 burstiness 5 burst 4 tfactor 0 21 31 station 4 generator d impulse cycle 3000 duty 0 15 burstiness 5 burst 4 tfactor 0 06 compute sum simplestats d_input compute sum simplestats d_trudelay An example FRACAS output Number of d input samples for station 0 is 299400 Statistics samples minimum maximum average var stddev 299400 0 328 48 04 2224 47 16 Number of d trudelay samples for station 0 is 14381736 Statistics samples minimum maximum average var stddev 14381736 272 312 21 WED Tee 13 6 32
28. is useful when the output is redirected to a program The v verbose option is useful for looking at the progress of the simulation FRACAS is a discrete time simulator Everything in the emulation happens at intervals of one frame and no measure makes sense in between In the input file the length of a frame is defined as an integer number of TMUs This is useful only to have all the times expressed in some handy time unit but does not affect the behaviour of the simulator in any way For example if a particular simulation uses a time frame 20 ms long it is useful to express all the times in milliseconds by setting the time frame to 20 TMUs This version of FRACAS does not allow a non integer time frame to be specified Moreover the time frame must be kept as small as possible if the packet by packet delay has to be computed This is because the maximum computable must fit into 16 bits 216 1 65535 In our example packet by packet delays of up to 65 seconds could thus be computed There is no limit on the delays if they are measured as a frame by frame mean delay Since all the timings are multiple of a frame no delay shorter than this quantization unit can be resolved This implies for example that if in the real system the delays are less than one frame long they are rounded down to zero and invisible to FRACAS In the input files keywords are used to define the network configuration the channel allocation policy and the statist
29. mpute the assignments The length of the assigned transmission window is proportional to the request in a range of values between a minimum and a maximum threshold mindall and maxdall The proportionality factor is proportional to the total number of stations in a range allnmin allnmax The complete expression for the assignment is bound p max a min b assignment boundmindall maxdall request boundalinmin alinmax stno alld where stno is the number of stations in the system Default values for allnmin allnmax and alld are 5 50 and 100 respectively meaning that the proportionality factor is bounded to a range of 5 to 50 by default After each assignment the datagram request is decreased by the assignment itself and the next request is analysed if space is still available in the frame The first assignment that does not fit entirely into the current frame is analysed as the first assignment in the next frame where the rest of the computed amount is assigned Any space available in the frame after an entire assignment cycle i e the time between two consecutive allocations to the same station is shared among all the active stations even those which had no datagram assignment in that frame 4 3 DRIFS This is a distributed control assignment algorithm 6 The allocation requests are broadcast to all the stations in the system therefore the delay between a request and the relevant assignment is about one
30. n s lt argument gt v lt argument gt d lt argument gt lt generator line gt lt generator line gt lt blank line gt generator line computer section lt computer line gt computer line ids generator slvld lt generator name gt lt generator arguments gt lt computer line gt computer lt ids gt lt computer name gt lt observable gt lt computer arguments gt lt number gt lt number gt sum all 29 30 REFERENCES N Celandroni E Ferro N James F Potorti FODA IBEA a flexible fade countermeasure system in user oriented networks International Journal of Satellite Communications Vol 10 N 6 pp 309 323 November December 1992 N Celandroni E Ferro F Potorti FODA IBEA TDMA System Description Final Report CNUCE Report C94 18 September 1994 N Celandroni E Ferro F Potorti The performance of a TDMA satellite system for non real time and real time traffic CNUCE Report C95 12 February 1995 N Celandroni E Ferro F Potorti Satellite bandwidth allocation schemes for VBR applications CNUCE Report C94 24 December 1994 N Celandroni E Ferro F Potorti M Conti E Gregori A bandwidth assignment algorithm on a satellite channel for VBR traffic submitted to the International Journal February 1995 N Celandroni E Ferro F Potorti Study of distributed algorithms for satellite capacity assignment in a mixed traffic and faded env
31. ness H rmdn seed burst where peakedness the ratio a between the variance and the mean of the samples per TMU The peakedness ar of the process over an interval of length T is ar a H the Hurst parameter of the process Usually in the range between 0 7 and 0 85 rmdn binary exponent of the length of a cluster of samples Default 18 5 6 5 5 External traffic generator external This generates TRUs at times specified in an external data file The file is in ASCII format with an integer number per line If the type is set to packet the numbers in the file are integers which represent the number of TRUs generated in successive frames The first line contains the number of TRUs generated in the first frame and so on one line per frame If the type is set to interval the numbers 21 are floating point numbers representing the distance in TMUs from one TRU generation to the next The first line contains the generation time of the first TRU the second the time elapsing from the generation of the first to that of the second and so on one line per TRU The times for the interval type can also be scaled usually to do time unit conversions by multiplying them by a constant indicated with the scale parameter An external process can feed the generator with the numbers instead of having them stored in a file In this case the parameter ipipe instead of ifile must be used This is the only generator that does not allo
32. or packet losses that trigger retransmissions in higher level protocols Datagram packets delivery is not guaranteed in the sense that the datagram transmissions may be momentarily suspended e g when congestion is detected and that packets exceeding the system buffering capacity may be dropped Moreover the delay a datagram packet experiences can occasionally be very high in the order of seconds By stream we mean connection oriented applications characterised by a constant packet arrival rate These applications typically require short and fairly constant delays They cannot tolerate out of order delivery of packets but can tolerate occasional bit errors and dropped packets In practice stream traffic needs a fixed amount of bandwidth and the satellite network should maintain a low and constant delay on the arrival of the information This kind of traffic is generated by applications that have constant throughput like voice slow scan TV fixed rate video conference measurement data and so on It is referred to as FBR Fixed Bit Rate traffic By VBR we mean Variable Bit Rate video which is one of the most interesting and challenging real time applications Constant quality VBR encoders attempt to keep the quality of video output constant resulting in a highly variable and bursty output bit rate VBR video traffic is both highly variable and delay sensitive so transmission over high speed networks for example ATM is generally implement
33. r excess division syntax allocator feeders na rbovh dquote bovh maxdall trace where na number of frames in which the allocations do not change assignment cycle length expressed in frames rbovh reference burst overhead in TRUs Is subtracted from the available space once every na frames dquote floating point number Fraction of the datagram request assigned in the datagram assignment cycle bovh burst transmission overhead in TRUs Must be less than framesize maxdall maximum datagram allocation in TRUs Must be greater than lt bovh gt Default is half the available space trace flag for enabling tracing y n Default n 5 4 4 DRIFS drifs The channel space is shared among the stations according to the DRIFS access scheme The stream allocations are equal to the sreg value of each station the VBR allocations are equal to the VBR request bounded by the vminreg and vmaxreq values of the vbrreq keyword the datagram allocations are computed using the DRIFS algorithm The delay between the request and the corresponding allocation is an integer number of frames equal to 2 rttime frametime where the division is an integer excess division syntax allocator drifs stinframe stovh dquote bovh maxdall trace where stinframe maximum number of control slots in a frame stovh control slot length in TRUs dquote floating point number Fraction of the datagram request assigned in
34. t do not have an explicit seed selection Zero means that the system time is used as the seed Default 0 max_cslen Maximum number of categories bins to use when the categorise computer is called with an unspecified number of bins Default 25000 12 ref_traffic Used as a reference traffic value by those generators which specify their mean throughput with a factor to be multiplied by this quantity rather than with an absolute number 5 2 INITER LINE Only one initer can be specified Since the allocators usually impose a delay of one or two rttime the first few frames have no requests The initer decides the allocations of these first frames 5 2 1 Equally distributed datagram allocation initer even The stream allocations are equal to the sreq value for each station the VBR allocations are equal to vminreq for each station the remaining available datagram space is evenly divided among all the stations syntax initer even 5 2 2 No datagram allocation initer zero Stream and VBR are allocated as in the even initer no allocations are given to the stations for the datagram traffic syntax initer zero 5 3 REQUESTER LINE Only one requester can be specified It computes the requests for each station It is called for each station once per frame 5 3 1 Queue requester queue For each type of traffic a request is issued equal to the current relevant queue length syntax requester queue 5
35. t observable gt ttygraph ofile lafile lopipe rows cols where 27 rows number of rows used for representing the character graph Default 22 cols number of columns used for representing the character graph Default 78 5 8 INPUT FILE SYNTAX A FRACAS input file is composed of a sequence of one or more run descriptions file contents lt run description gt lt blank line gt end lt blank line gt lt run description gt run description lt global variable section gt lt initer section gt lt requester section gt lt allocator section gt lt stopper section gt lt station section gt lt station section gt lt computer section gt global variable section lt global variable line gt lt global variable line gt lt blank line gt global variable line global variable name gt lt global variable argument initer section initer lt initer name gt lt initer arguments gt lt blank line gt requester section requester lt requester name lt requester arguments gt lt blank line gt allocator section allocator lt allocator name gt lt allocator arguments gt lt blank line gt 28 stopper section lt blank line gt station section stopper lt stopper name gt lt stopper arguments gt station lt number gt lt number gt streamreq sreq lt argument gt vbrreq vminreq lt argument gt vmaxreq lt argument gt maxqueuele
36. t printed with the terse option syntax computer lt stations gt lt observable gt simplestats ofile lafile lopipe 5 7 2 2 Quantile computer quantile This computer prints the quantile of an observable The 0 9 quantile of an observable is the sample such that 90 of samples are less or equal than it Each 25 output line contains the requested quantile and the result preceded by a header line and followed by a blank line The header and the blank line are not printed with the terse option syntax computer lt stations gt lt observable gt quantile ofile lafile lopipe q where q the fraction of the number of samples less or equal to the value printed by the quantile computer It is a floating point number in the 0 1 interval 5 7 2 3 Sample listing computer listing This computer lists the values of all the samples observed One parameter every and three styles are available The normal style prints one value per line taking one sample every every samples The aggregate style prints one value per line which is the sum of every successive non overlapping samples The compact style prints two values per line It groups successive equal samples in one line the first number printed is the number of equal successive samples the second number is their value The parameter every makes no sense with this style The listing of the values is preceded by a header line and followed by a blank
37. tiness which is the ratio between the mean traffic rate instate 1 and the mean traffic overall traffic rate The ratio of the traffic rate in state 1 to the burstiness burstiness duty 1 burstiness duty traffic rate in state 2 is expressed by ratiol2 ratiol2 1 duty ratio12 duty conversely burstiness seed is an integer All the other parameters are floating point values syntax generator traffic type impulse traffic ltfactor cycle duty burstiness seed burst where cycle the mean cycle length length of state 1 plus length of state 2 in TMUs duty the ratio between the mean length of state 1 and the mean cycle length burstiness the ratio between the mean traffic rate of state 1 and the overall mean traffic rate the one indicated by the trafficltfactor parameter 5 6 5 4 Fractional Gaussian Noise traffic generator fgn The number of TRUs generated per frame by this generator approximates a Fractional Gaussian Noise fgn process The generator uses the Random Midpoint Displacement rmd algorithm to approximate the fgn process The algorithm generates a cluster of pow 2 rmdn samples at a time If the length of the simulation is greater than this value a new cluster of samples is generated not correlated with the previous ones The mean of each cluster of samples is forced equal to the traffic requested syntax generator traffic type fgn traffic ltfactor peaked
38. w a trafficltfactor parameter syntax generator lt traffic_type gt external ifile lipipe type scale burst where ifile name of the file containing the numbers in ASCII format one per line ipipe the command line used to invoke a process which writes on its standard output a sequence of ASCII numbers one per line type the string packet or the string interval scale floating point number that multiplies the numbers read from the file or pipe Only allowed with the interval type Default 1 5 7 COMPUTER LINES A computer is a routine that operates on observables Each computer makes a different operation Computers can be added to FRACAS 22 5 7 1 The observables Observables have names starting with s_ or v_ or d_ These prefixes stand for stream VBR and datagram respectively meaning that the statistics collected only relate to the respective type of traffic Here is a list of the available observables and a brief discussion for each of them s_input v_input d_input Number of TRUs presented to the input of a station per frame The TRUs are generated by the generators declared in the station blocks These numbers are the sum of those generated by the station s generators at each frame For the pseudo station sum the observable is the sum of the inputs to all the stations s_queue v_queue d_queue Length of the input queue of a station Each station receives the TRUs generated by its gen
Download Pdf Manuals
Related Search