Home

PILOT/CICS User Manual

1. TOTAL RUN TIME sec 180 TIME TERVAL sec 01 OF TERVALS PER REPORT 100 RIPTION 4381 2 D mips 2 6 OF CPUs 1 OF WLs 4 K SERVICE TIME sec 044 E PACK BUSY 128 KKK K KKK A RESULTS KKK K KKK A TIME USED sec 180 00 CPU 1 92 42 Workload X Count CPU Page Flts CPUR 1 SYS OVHD 20 0 49 0 0424 2 CICS1 320 67 01 377 3758 3 CICS2 414 24 90 175 1080 4 CICS3 24 0 01 2 0008 F9 PRINT RESULTS ETURN TO START Figure 13 Simulator Intermediate or Final Report Notes 1 Time Used The total run time CPUn The CPU utilization of each processor When modeling multi processors the percent busy may not be even Extensive design in the model attempts to even out the usage across all processors When modeling light loads the first processor may show a higher value when the model execution time is very short The actual percent busy on a live system would be even due to the micro code in multi proces sors 2 Workload The workload and its priority X Count The number of requests or transactions processed If X Count is divided by the time used the result should equal the arrival rate PILOT CICS 49 KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc November 1 1999 CPU The percentage of all processors kept busy to execute the workload Page Flts The total number of page faults for the
2. WL The name operand is the Workload name pg is the Performance Group number s describing this workload OTHER SIMBUILD will create a separate work load to combine all workloads not defined when OTHER Y All performance groups not previously used will be combined in a workload called OTHER This operand can be speci fied on any of the control cards but will only create one workload PRTY The dispatching priority 1 20 of this task Two or more workloads can have the same priority APPLID VTAM applid used to identify CICS regions when SMF 110 records are input into the model generating program The four character TCT id is used to identify CICS regions when The Monitor records are used for input DISKS Where voll vol2 are the disk packs used by the workload Disk pack names can be generic by specifying an This parameter is used to calculate the average device service time and average device utilization PEAK TIME HHMM hhmm WKDAYS Y N where PEAK Is the control card keyword KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc TIME The start and end time of the peak period being processed across the entire input file This parameter is optional WKDAYS This parameter controls the inclusion of weekends to average the values across the input file The default is Y meaning to omit weekend data SIMBUILD JCL The example below shows the cr
3. Attempting to change a protected field will produce a beep as an error indication The auto recalculation switch has been set off so be sure to press the EN after all changes This will recalculate all the dependent cells Above each set of input cells is an unprotected cell called TITLE which will appear as a title of the graph created for this option GRAPH Produces a graph of the results calculated by CHANGES These graphs will differ depending on the option chosen The GRAPH option requires a graphic display and adapter See the Lotus 1 2 3 manual for a description of the supported displays To save the graph for a printer or plotter output type Sains save and give it a name as described in the Lotus 1 2 3 manual Once a graph has been selected it can be ge again by simply pressing the NH GRAPH key If a model parameter is Recalculate or Calc key graph KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc November 1 1999 changed first press S83 to recalculate and then press S83 to graph The new results will be displayed Using the Spreadsheet This section describes how to use the options provided in the MODLCICS spreadsheet Using the Response option o Invoke the user menu by pressing the valle Aa keys e Press EM to select the Response option e Press MM to select the Change option e Change one or more model input param eters The XR value is the st
4. D1 D2 or D3 TIME Specifies the time of day in hhmm for mat CID CICS system identifier specified in DFHSIT 1 7 DFHTCT 2 1 SYSIDNT parameter TASK The number of transactions processed for this period AVG RSP The average response time for all trans actions AVG WAIT The average wait time for all transac tions AVG CPU The average CPU time for all transac tions STOR The amount of operating system storage available OSCOR in 1024 byte units AMC The total number of access method calls for all transactions I O The total amount of I O processing all transactions performed This includes access method calls journal puts synchpoints BMS In BMS Out Tempo rary Storage AUX count and Transient data get and put counts KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc CPU CICS The total percentage of the CPU used by the CICS region Sum of TCB and SRB CPU time TMON collection option REALCPU must be set to Y CPU APPL The percentage of the CPU used by the application programs running within the CICS region TMON collection option REALCPU must be set to Y PAG IN The page in rate pages per second for the CICS region PAG OUT The page out rate pages per second for the CICS region COMPS The number of program subpool com pressions during the period PCP_RSP The average time for loading a program during the period PCP T
5. Adding Real Memory Adding real memory will affect CICS re sponse time in proportion to the page fault rate In other words if the page fault rate is low i e 1 to 2 pages per second then the effect of adding real memory will be minimal If the page fault rate is high the addition of real memory can have a major impact on performance Since you typically November 1 1999 add real memory in large blocks 4 or more megabytes one can assume that page fault rates will be dramatically reduced If the current rate is 5 to 7 page faults per second adding 4 or more megabytes of memory will initially reduce the page fault rate to 2 or less Changing the DASD Subsystem Changes to the DASD subsystem can be modeled by changing the following model input parameters page pack service time and utilization TP amp UP and data base pack service time and utilization TD amp UD DASD changes can include adding additional disk control units channels disk packs or cache controllers Choosing model input values to model these changes will require intelligent guestimates based on a good understanding of the differences between the current DASD subsystem and the proposed changes As an example assume the current subsystem consists of 3350 disks and that you wish to model the effects of converting to 3380s Also assume the base value of TD UD TP and UP are 035 40 030 and 36 respectively With the knowledge that aver
6. DASD can be done by changing the average service times and utilization of the devices of the workloads effected by the change Do not change the I O rate of the workload this can skew your results Do not forget to change the paging subsystem s parameters if it is upgraded as well The methodology is similar to that of the paging subsystem mentioned in the last section In 42 PILOT User s Guide November 1 1999 general when upgrading to a faster device the service times can be reduced If a device is being changed from a single density to double density the utilizations and service times will probably increase due to the nature of the architecture If two single density devices with utilizations of 20 used in random access online data bases are combined to a double density drive the utilization will more than double There will be a significant increase in the amount of arm movement over the previous configuration These are some of the factors one must consider when modeling DASD KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc Introduction The program SIMBUILD will generate a model for the simulator SIMCICS and the input parameters for the analytic model MODLCICS The model generator will use control cards RMF CMF and The Monitor records as input The output of the model generator must be down loaded to the PC in order for SIMCICS to process it Additional input is needed for records n
7. Is the control card keyword If this con trol card is omitted calculations based upon processor speed will be invalid Other fields not effected by speed will be correct Only the model name is needed if the processor is included in the PILOT s internal table MODEL Is the name of the CPU model i e 4341 12 3090 200 3084 QX etc Upto 8 characters can be specified A list of PILOT CICS 43 KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc November 1 1999 the models in the supplied table is provided in the Techniques section of this User s Guide When you use the name of a supplied model the other parameters are obtained from the table automatically SPEED The processor MSU value used by the SRM PROCS The number of processors for the model specified MIPS Millions of instructions per second for this processor ICS The suffix value for the IEAICSxx member to be processed This member is read from the PARMLIB DD statement and is used to relate RPGN Report Performance Group Numbers to the associated CPGN Control Performance Group Numbers when the workloads are characterized WKLD WL name pg OTHER Y N APPLID aaaaaaaa PRTY nn DISKS voll vol2 where WKLD Is the control card keyword If this control card is omitted a workload will be defined for each control performance group Up to 1 000 workloads may be defined for analysis 44 PILOT User s Guide
8. PMD sy a be ek he he ane BO Se i Eg 2 10 43 SUMA tb GS Foca e a hg ES E aes 1 3 5 9 11 14 18 19 SYSED cyte ee ooo safe wee oe ae 3 1 9 11 13 17 18 20 22 24 26 29 30 33 39 47 TASKS Fake A ns Woe AM a ee ete 5 14 18 26 39 mplale eos a ao we Ge dew aut eg hgh Beate Ste Shah ESA 5 7 14 29 31 HEMOS eG ae che nee oe untae nies Bad as i da 41 42 DEDE ob jee SS oe ee Ge eee Th 2 5 7 11 14 17 27 29 30 33 43 45 47 48 transactions ose A 1 4 9 12 17 19 22 27 33 35 36 38 39 48 Ype maay a A a q A A 1 20 24 33 43 52 PILOT User s Guide KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc November 1 1999 ATE sts Ahh A amp G aut axe te des Sagan SR 2 10 43 WU ZA OM remar 4 Gch eos ee oS Stata es oe E ee a 17 18 24 26 33 38 40 42 47 VOM Lui a GM pein ee ek ee ac R eee a ake Ee et 34 38 Workload ei A o a down ae ae ic Paty ew Dis aid 18 33 36 38 42 46 48 PILOT CICS 53
9. Satie Baek Ge ae a 18 24 34 39 40 A A at UR Gey una 1 2 9 10 17 19 27 29 31 33 35 43 parmlib ss ah nce Be A ses RS ae ee ete Be ee as way ae ae tee 42 43 ACETIC ade Sib aw ta dy els bate ane eee 35 37 performante a gs 3 1 3 9 11 17 18 22 25 31 33 35 39 42 43 performante UNE ora aa a adas 1 9 PUOUCICS prats aa gs eee 1 1 5 7 9 10 14 17 19 22 36 38 41 45 47 A tar xeon de UE O inte Ty a e 18 22 36 38 41 POSTO a oe ee Bee A teeta Ne GAS Mt eh ae Re eg Sa Se 43 pint s eGo price acta E De SP Da 5 7 14 29 31 47 PU oia 2 5 7 10 14 sI A O AN 9 TESPOMSE ut O y 1 4 7 11 14 17 19 25 27 29 31 33 37 39 48 response time sus air e 2 4 7 11 14 17 19 25 27 29 30 37 38 48 O O 18 22 36 37 41 43 AA AI A Tense deb 19 21 24 43 EI adh e O a o is tee ee E 2 10 SEUD sec orais E ES bosons Seana ean 5 7 14 29 A A a 19 35 36 41 43 45 47 SIM CICS 385 sr sta dd 17 33 38 41 43 45 47 SR EN 17 33 41 45 47 Mer a a ho ed a da od 1 2 9 42 43 Mic A A Bee Tae Se oe A a TT tan DEA 1 2 9 11 SVG TN iss So tye ad cept fascia aU Se eo SEs Tete saved Fang c o de dele aula ele leaf th ias ad 1 2 9 43 SIUM OS is ua et arate ww eee a an eter Mae ace ete sete a ned 1 2 9 10 STC LL ca ar o e ce eas Seg bse at Bad Sark wane se oe dR Gee Aes LS 12 STUN OCU a5 RS AS A A Nc 9 10 Spreadshirt desd es ts dE is 3 5 7 11 12 14 17 22 29 30 Md ee ace ae See reia tr 1 9 10 23 33 41 43 47 SUC
10. can be identified by their performance group number The RMF reported CPU service times can be converted to CPU seconds by using the appropriate conversion factor Dividing the CPU seconds used by the elapsed time yields the utilization by higher priority tasks In addition the utilization of the operating system which has not been included in a reported performance group must be added to this number This can be esti mated by comparing the total CPU utilization reported by RMF with the total number of CPU seconds reported for all performance groups The difference between the total utilization and the total reported utilization for all performance groups is the unreported utilization charged to system overhead PFR KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc The average page fault rate for the CICS task Expressed as page faults per sec ond this number represents the number of times each second that CICS must stop to wait for a page fault to be resolved This value can be obtained from the PILOT CICS summary spreadsheet XR The average CICS transaction rate ex pressed as transactions per second This value is obtained from the PILOT CICS summary spreadsheet or it can be calcu lated by dividing total number of CICS completed transactions by the elapsed time The input parameters can be produced by the model generator program SIMBUILD Depending on the options selected the user may chang
11. capacity Unless you can answer these questions you will find yourself putting out fires instead of planning to meet your company s business needs with data processing solutions Identify Resource Utilization by Business Usage Forecasting existing workload growth and its impact on current hardware configurations is the most time consuming process of forecast analysis There are two approaches you can take First study historical data and look for patterns in workload volume growth If you have the data look at what happened when a new on line application was implemented Did the transaction rate go up How much Similarly look at the effects caused by adding new users or new terminals for existing applications Historical data is often the most reliable measure of predicting future trends Second talk to end users development groups and strategic planners for future business needs refer to technique section of this manual The latter approach is more time consuming but may yield better results The first approach can yield good results if historical data is maintained and you are familiar with the company s growth patterns A combination of the two KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc approaches may be used to obtain good results PILOT CICS can be used to analyze data from a year ago to be compared to the volume usage today Also notice the resource consumption of the system at th
12. for the transaction rate equal to the average mea sured XR for the peak hours of CICS activi ty Project the transaction rate growth based on your installation s plans These plans should include the following e New CICS applications e Growth in the use of existing applications due to new users and e Growth in the use of existing applications due to increased activity by existing users If CICS usage data has been tracked this data can be used to help in your projections Use the BEST WORST option to calculate CICS response times for the twelve months of projected transaction rates Model input parameters for best case should be set to the base values currently measured in your installation Worst case values should be set to reflect possible growth in one or more model parameters over the next twelve months For example if the page fault rate varies from 2 to 4 pages per second on a daily basis use the following technique Set PILOT CICS 25 KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc the best case value of page fault rate PFR to 2 and set the worst case value to 4 This assumes that over time the average PFR may grow from 2 to 4 pages per second This approach can be used with other model parameters as well Be sure to input the maximum acceptable response time MAX RSP calculated in step 2 When a graph of BEST WORST is displayed the response times are plotted over the twelv
13. hours is not going to give an accurate picture of the system s behavior during times of peak usage Since most on line systems show an umeven pattern of demand and performance it would not be productive to average these peaks and valleys over 24 hours Most systems are at their peak usage only once or twice a day and these periods are generally an hour or less in duration If you choose to model a long period of time where resource consump tion is not consistent then some addi tional analysis must be done to account for the peak periods during that time frame Sometimes it is useful to see how a system will run during non peak times as well as during peak periods If this is required then two models should be used November 1 1999 and analyzed separately Peak period time frames can be identified through the use of PILOT MVS and PILOT CICS If PILOT MVS was not purchased then the data needed can be obtained from RMF or other tools that provide similar re porting capabilities Once your peak periods have been identified you need to measure average resource consumption during these periods At least 5 days a week of data preferably 20 days a month should be summarized to get this data Tracking Data Much of the data required as input to the simulator is reported by PILOT CICS By analyzing trends and variations in this data over several peak periods you will be better able to choose the proper input PILOT CICS reports
14. misplaced decimal point or wrong use of units i e transactions per minute instead of per second will certainly invalidate the model Another cause may be an internal bottleneck within CICS Internal bottlenecks represent conditions which degrade CICS performance despite the availability of CPU cycles memory and DASD resources Some examples are improperly set values of MAXTASK VSAM buffers and IMS DATABASE strings These represent artificial constraints to performance and cannot be accounted for in the model Ina sense these bottlenecks represent tuning problems and must be separated from capac ity planning issues The process of calibra tion must be completed before forecast analysis can be done Forecasting Future Hardware Re quirements After the model is calibrated you are ready to play the What if game There are two 40 PILOT User s Guide November 1 1999 categories of questions capacity planners are most often asked They are direct hardware questions easiest to answer and application growth questions This section will deal with hardware questions Refer to the techniques section for application growth questions Regardless of the initial question all capacity planning issues reduce to the following questions What impact will this have on my current environment Will this affect response time and By how much and finally How much longer will it be before I will be out of
15. onon AUNA Model Input Parameters CPUR CPU Service Time UP Page Busy UH Super Overhead TP Page Service Tim UD Data Busy PFR Page Fault Rate TD Data Service Time IOR I O Rate XR Arrival Rate TITLE Current System Delta of Arrival Rate 0 1 o IA UF WU HO TP UP TD UD IOR UH PER 0 035 0 150 0 035 0 300 5 000 0 200 2 2 to 20 23 Nov 93 03 31 PM Figure 6 MODLCICS Template Response Panel PILOT CICS 31 KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc November 1 1999 Al w10 ENU HANGE GRAPH SAVE_GRAPH PRINT RETURN hange one or more model input parameters Press CALC ALT F for the menu A B E D F G H L Model Spreadsh Spreadsheet Version MODELCIC v1L5 0 The following short cuts are available ALT M This Menu ALT B Best Case Worst Case ALT R Response Time ALT O Option Comparison ALT F Sensitivity Factors ALT C CPU MIPS Conversion Use CTRL x in the Windows version of Lotus O CO JU uN p Model Input Parameters CPUR CPU Service Time UP Page Busy UH Super Overhead TP Page Service Tim UD Data Busy PFR Page Fault Rate TD Data Service Time TOR I O Rate XR Arrival Rate TITLE Current System Delta of Arrival Rate Oie TP UP TD UD IOR UH PFR 0 035 0 150 0 035 0 300 5 000 0 200 2 2 I I I I I I H H H H t f k H H 0 JU UN R O to 20 23 Nov 93 03 32 PM igure LCS Template Fa
16. possible on a dyadic processor will be something less than twice the transaction rate calculated for a single copy of CICS The degree of asymmetry between the 2 copies of CICS as well as the overhead introduced by MRO will tend to reduce the factor to a value less than one half If an upgrade to a multi processor 2 processors is to be considered and if it is expected to run 2 copies of CICS calculating the maximum transaction rate possible with one copy of CICS is done by multiplying the transaction rate by 2 and then reducing this number by a factor of half the original transaction rate As an example assume the transaction rate on a single processor is 18 transactions per second Assume an upgrade to add a processor with the same MIP speed The maximum transaction rate possible is 18 2 18 2 25 transactions per second Model Description The following equations are used to calculate the average CICS response time RSP as a function of the 9 independent parameters in the CICS analytic queuing model These parameters are described in Model Parameters RSP CPUR PTD UGT JOR E AAA DO AA A OD 1 UGE 1 UGE 1 UGT PILOT CICS 29 KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc November 1 1999 where IODT IOR Esc o TD 1 UD PDT 75 up _ TP 1 UP UGT UH PFR PDT UH PFR PDT UGE XR CPUR UGT 30 PILOT User s Guide KLM Technical Speci
17. program PCFMT EXEC PGM SMFPCLMK REGION 1720K or PCFMT EXEC PGM MONPCLMK REGION 1720K STEPLIB DD DSN TSUO0l1 MYLIB DISP SHR LMKIN DD DSN BACKUP LANDMAR K DATA G0001V00 UNIT TAPE VOL SER 123456 DISP O LD ISMFLOG DD SYSOUT A ISMFMONS DD DSN TSU01 CICSS UM PRN 1 DISP CATLG 7 SPACE TRK 4 4 RLSE ISMEMONSR DD DSN TSUOI CICSRSP PR N PILOT CICS 11 KLM Technical Specialties Inc Exclusive Distributors if DISP CATLG if SPACE TRK 4 4 RLSE ISMFCTL DD x COLLECT CICS TRANS EXCEPT SIGNON SIGNOFF TRANS REGION EXCL CSSN CSSF IREG PRD1 Notes 1 STEPLIB Used if SMFPCLMK is not in the link list 2 LMKIN The input data set containing TMON re cords 3 SMFLOG Statistics on the run 4 SORTIN Used as a workfile to calculate percen tiles 5 SMFMONS CICS summary file to transfer to a PC 6 SMFMONSR The CICS summary response time file to transfer to a PC 7 SMFCTL Control card data set Spreadsheets There are two spreadsheets available for the analysis of CICS performance data The Summary spreadsheet data is obtained from the file created in the SMFMONS file and is illustrated in Figure 3 on page 14 The following information is provided in the SMFMONS file for each hour DATE 12 PILOT User s Guide Axios Products Inc November 1 1999 Specifies the date CICS was executing The format can be mm dd yy or Lotus 1 2 3 date format
18. transaction TASK The number of transactions processed for this period 95 RSP The 95th percentile of the transaction response time for this transaction code This means the value reported is higher than 95 percent of the transactions pro cessed for this period 75 RSP The 75th percentile of the transaction response time for this transaction code 50 RSP The 50th percentile of the transaction response time for this transaction code 25 RSP The 25th percentile of the transaction response time for this transaction code MIN RSP The minimum transaction response time for this transaction code MAX RSP The maximum transaction response time for this transcation code 95 WT The file control wait time of the 95th percentile value as described above 50 WT The file control wait time of the 50th percentile value as described above 14 PILOT User s Guide November 1 1999 MIN WT The minimum file control wait time for this transaction code MAX WT The maximum file control wait time for this transaction code CPU The average amount of CPU time the transaction used STOR The average amount of storage the trans action used in 1024 byte units AVG I O The average amount of I O processing the transaction performed This includes access method calls journal puts synchpoints BMS In BMS Out Tempo rary Storage AUX count and Transient Data get and put counts DLI RSP The average response time f
19. 2 call if DB2 is installed This field is not available in CMF This field is provided for compatibility with TMON 8 2 DB2 The number of DB2 calls during the period only for the DB2 option This field is not available in CMF This field is provided for compatibility with TMON 8 2 UDB RSP The average response time for a User Database This field is not available in CMF This field is provided for compat ibility with TMON 8 2 UDB The number of User Database calls during the period This field is not available in CMF This field is provided for com patibility with TMON 8 2 The Response spreadsheet data is obtained from the file created in the SMF110R file and is illustrated in Figure 2 on page 7 The following information is provided in the SMFPC110R file for each transaction pro cessed 4 PILOT User s Guide November 1 1999 DATE Specifies the date the transaction was initiated by CICS The format can be mm dd yy or Lotus 1 2 3 date format D1 D2 or D3 TIME Specifies the time of day the transaction was initiated by CICS in hhmm format TRAN The transaction name T Transaction types as follows A Auto initiated transaction P Printer transaction T Terminal initiated transaction Z Conversational transaction TASK The number of transactions processed for this period 95 RSP The 95th percentile of the transaction response time for this transaction code This means that the value report
20. 8 11268 9514 14 07 14 93 2300 TS04 192 253 143 1019 11268 1994 15 07 15 93 PS04 218 875 746 0192 1 952 277 16 07 15 93 100 TS04 183 SE 065 0173 1 6744 850 17 07 15 93 200 TS04 737 201 064 0488 1 7404 5344 18 07 15 93 300 TS04 1925 136 017 0151 alt 7536 12343 19 07 15 93 400 TS04 284 971 746 1 9180 5208 20 07 15 93 500 TS04 339 801 0 606 9180 6190 23 Nov 93 03 18 PM CMD Figure 3 Summary Template for The Monitor 16 PILOT User s Guide KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc November 1 1999 Al MENU IMPORT GRAPHS PRINT RETURN SAVE S EXIT Load a PRN file for a CICS Response Ti Report A B C D E G I Spreadsheet Version RSPCICS v1L5 0 CICS Response Time Templat PILOT CICS C 1986 1993 1 2 3 4 5 6 7 8 9 There are macros you can use to return to the menus Press T A to bring up the query menu G to bring up the graphs menu M to bring up the main menu this screen P to bring up the print menu Use CTRL in the Windows version of Lotus A A A A o ITAU E WU HO to 20 23 Nov 93 03 20 PM Figure 4 Response Template for The Monitor PILOT CICS 17 KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc November 1 1999 18 PILOT User s Guide KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc Introducti
21. B DISKS IMS0 MFG INV PEAK TIME 1400 1600 Job card EXEC card with the program name NO Re 3 STEPLIB provides the name of the PILOT load library if it is not in the system s LINKLIST 4 SMFIN is the input file for RMF type 70 75 and CICS CMF SMF type 110 PILOT CICS 45 KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc records The RMF records must be concatenated before the CICS records MONIN is the input file for monitor data which is not ina SMF format This file can be in compressed or uncompressed format PARMLIB is the library that contains the ICS identified by the ICS control card parameter The member IEAICSxx may be stored in any library SYS1 PARMLIB is only used here as an example SIMOUT is the output file formatted for SIMCICS on the PC This file must be downloaded to the PC SIMLOG contains the input parameters for the analytic model MODLCICS There is no file to download Just enter the nine numbers as shown SIMCTL is the control file for input parameters 46 PILOT User s Guide November 1 1999 KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc November 1 1999 Appendix A The following is a sample of the Simulator s Parameter Selection Screen PILOT SIMCICS PARAM ECTION SCR CONFIGURATION FILENAME CONFIG TOTAL RUN TIME sec 180 TIM
22. E INTERVAL sec 0100 OF INTERVALS PER R 100 CPU DESCRIPTION L 2 CPU SPEED mips a OF CPUs PAGI K SERVIC E sec 0440 PILOT CICS SIMCICS MOD Axios Products Inc TITLE FO THIS RUN PILOT CICS SIMBUILD MODEL ERATOR OF WLs 4 E PACK BUSY 1280 PF1 HELP PF2 NEXT PF3 SAVI PGM Figure 11 Simulator Parameter Selection Screen PF5 CPU TABLI PILOT CICS 47 KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc November 1 1999 The following is a sample of the Simulator s Workload Definition Screen PILOT SIMCICS WORKLOAD DEFINITION SCREEN PATH ARRIV I O PF DSK DSK OF S single task WORKLOAD PRTY LENGTH RAT ERV BUSY TERMINALS 1 SYS OVHD 1 101520 z 0000 0000 0 ETESA 908673 027 6 1853 100 2 2 3 GCTCS2 2 288850 z 0286 2553 40 4 3 CICS3 3030 0225 1075 20 ELP F2 BACK F3 SAVI E F5 CPU TABLE F6 RUN F10 END PGM Figure 12 Simulator Workload Definition Screen 48 PILOT User s Guide KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc November 1 1999 The following is a sample of the Simulator report The format of the intermediate and final reports is the same PILOT CICS SIMCICS TITL HIS RUN PILOT CICS SIMBUILD MODEL ERATOR
23. LD program from the control cards entered or created by the SIMBUILD program from the Control Performance Group number PRTY The dispatching priority assigned to a workload 1 20 with 1 being the first to be dispatched PATH LNTH The average number of instructions re quired per transaction or service request ARRIV RATE The arrival rate of transactions or re quests for service The following parameters account for the delays a workload may encounter while processing a transaction or request NO RATE The average number of EXCPs complet ed per transaction or request PF RATE The average page fault rate of the work load in pages per second DISK SRV November 1 1999 The average disk service time in sec onds for the workload DISK BUSY The average percent disk busy for the workload TERMINALS The maximum number of terminals for a workload This parameter when specified will cause the simulator to act like a closed queuing system If the de fault of zero is specified the simulator will act like an open queuing system The simulator generates transactions according to the arrival rate specified If the system being modeled becomes constrained the number of requests on the queue for that workload will continue to grow as the transaction rate is maintained For example if a workload is defined with a FTERMINALS set to 100 the simulator will stop generating new transactions when the number trans a
24. Mid p Nhe Ah A A E an 3 PID T CICS Tor The Montoro Ga os 9 AA ta erage es Sy RRR Ge FG ae iy PR E SE ae hale Hee 9 Control Cards and Parameters 2 fia bas oe E A eee ER A a S 9 Parameters on the JCL EXEC Card i635 46 ae Oe bee a Ge 10 E Sea cay ahaa cs AS Ge DANA eats Sp errr Ec 10 ji MA II EN 11 Modelos naaa re ro Bae Sah BAe Ga te Dole b yg edhe 17 introduction t Modeling sas ESA eA Gee oN a es 17 NODC KS Spreausneet 3 x orador tai Gilg that wd Oe Ae oo actos 17 Model Paramore 17 Spreadsheet Features iS Aia AR A RD Pl 19 Ops A IS At easy einen Gomes g 19 Usine the Spreadsneet 2 andare oe Awad oo tye er hal elke se 6 Se 21 Methods eck ete o Ratatat to sl ails 22 Calibrating the Model sossegada s pan Bs Ga END DES dais 22 Calculate the Maximum Acceptable Response Time 0000 23 Create a Capacity Plan Based On Expected Transaction Rate Growth 23 Identify Performance Bottlenecks uc a A we Oe PR GS 24 Compare Configurations to Solve Capacity Problems 44 24 Choosing the Model Input Parameters o 0 00 0 25 Adding Real Memory umano E oe Gd Bw seas in hk EPO GP Bead sitet a 25 Changing the DASD SUD SM EE Rg A RES ORS E A ees 25 Changing the CPU aa gta ek SO a BS Ta at Di det A eee E e 25 Model Descrplla a ts aire peta AS Gor a E AR uae hee 21 SEM CWS SHADE a a A a ee tet Reng yt a 33 STM TOCUCIIOM A E RN 33 Parametros IE EAS NS Ad 33 Parameter S lection Sereen rsa tea riada a
25. NUNNNNNN WN Da 23 Nov 93 03 35 PM H H Figure 9 MODLCICS Template Options Panel PILOT CICS 33 KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc November 1 1999 W8 TABL E CPU PERFORMANCI CHANGE_ VALUE MODEL1 EL2 OTHER RETURN the CPU table of values Press ALT C to return to this menu LE CD CE CE CG CH CI CJ CK E OF RELATIVE CPU PERFORMANCE LTI PROCESSORS ARE RATED AT SINGLE PROCESSOR VALUES help convert CPU Seconds on existing CPU to new CPU CONVERT CPUR1 ON CPU1 TO CPUR2 ON CPU2 AS FOLLOWS CPUR2 CPUR1XMIPS1 MIPS2 CPU MODEL After selecting the values choose Other then Hit CALC KEY F9 Then hit ALT C 3031 3031 AP 3032 3033 AP 3033 MP 3033 N 3033 S 3033 U 3033 U 3081 D 3081 G 3081 GX 3081 K 23 Nov 93 03 37 PM CPUR1 MIPS1 MIPS2 0 220 2 84 4 03 UuNnnonKRR RYE BP Figure 10 MODLCICS MIPS Conversion Panel 34 PILOT User s Guide KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc Introduction The simulator SIMCICS provides an innova tive and interactive tool for capacity planning and performance analysis of complex online environments The simulator basically works dispatches workloads in much the same matter as a multi tasking operating system Workloads such as CICS
26. PILOT CICS Axios Products Inc 1373 10 Veterans Highway Hauppauge NY 11788 3047 Sales Administration 800 877 0990 Technical Support 516 979 0100 Telecopier Fax 516 979 0537 Preface This publication contains information necessary for the operation of PILOT a family of proprietary program products used for performance management and capacity planning It provides data processing managers system programmers and capacity planners with information required to use this product Information in this publication is subject to significant change THIS MANUAL IS PROVIDED FOR THE SOLE AND EXCLUSIVE USE OF THE CUSTOMER THE MATERIAL CONTAINED IN THIS MANUAL IS CONFIDENTIAL AND SHOULD BE SO TREATED COPIES MAY BE PURCHASED FROM AXIOS PRODUCTS INC ANY UNAUTHORIZED REPRODUCTION OF THIS MANUAL IS PROHIBITED Sixth Edition November 1999 This edition applies to Version 1 7 of the PILOT program products and to all subsequent versions and modifications until otherwise indicated in new editions or newsletters Copyright 1987 1999 KLM Technical Specialties Inc All rights reserved Axios Products Inc exclusive distributor KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc November 1 1999 Contents PID CICS F r MiB id nadaa Gul ay ae pa 1 ATM TOGUC OA AS A A A e RR CA 1 Control Cards and Parameters o bi ei as 1 Parameterssonihe JOEL EXEC Card id ES Sa 2 A ER 2 Spreadsheets uid cute E O
27. Specialties Inc Exclusive Distributors Axios Products Inc November 1 1999 Al W9 MENU IMPORT GRAPHS PRINT RETURN SAVE SETUP Load a PRN file for a CICS Summary Report A B E D E F G H T J Spreadsheet Version SUMCICS v1L5 0 1 2 3 4 PILOT CICS Summary Template 5 6 7 AXIOS PRODUCTS INC PILOT CICS C 1993 TMV8 KLM Technical Specialti CICS REGIONS TS04 DATE TIME CID TASKS AVG_RSP AVG_WAIT AVG_CPU OSTOR DSA AMC 07 14 93 1900 TS04 T29 0 089 0 038 0 0124 678 11268 1 07 14 93 2000 TS04 184 055 022 0 0055 678 11268 109 1 07 14 93 2100 TS04 300 093 023 0092 678 11268 1334 13 07 14 93 2200 TS04 403 295 058 9881 678 11268 9514 14 07 14 93 2300 TS04 192 253 143 1019 11268 1994 15 07 15 93 PS04 218 875 746 0192 1 952 277 16 07 15 93 100 TS04 183 SE 065 0173 1 6744 850 17 07 15 93 200 TS04 737 201 064 0488 1 7404 5344 18 07 15 93 300 TS04 1925 136 017 0151 alt 7536 12343 19 07 15 93 400 TS04 284 971 746 1 9180 5208 20 07 15 93 500 TS04 339 801 0 606 9180 6190 23 Nov 93 03 18 PM CMD Figure 1 Summary Template for CMF 6 PILOT User s Guide KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc November 1 1999 Al MENU IMPORT GRAPHS PRINT RETURN SAVE S EXIT Load a PRN file for a CICS Response Ti Report A B C D E G I Spreadsheet Version RSPCICS v1L5 0 CICS Resp
28. age seek times should improve from 025 to 015 seconds and the data transfer rates improve from 1 2 to 3 0 megabytes per second 3350 vs 3380 the estimated new parameters are follows TD from 035 to 020 UD 40 to 25 TP 030 to 018 and UP 36 to 20 These reductions are estimates based on both a knowledge of the current environment as well as an under standing of the proposed changes In most installations the 4 DASD parameters will not have a large impact on performance and errors in these estimates will have a small PILOT CICS 27 KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc effect on calculated response time If the impact of a DASD change is uncertain assume a conservative decrease in these parameters and examine the sensitivity factors to measure their impact on response time If the sensitivity factors are low then the effect will be minimal Changing the CPU Probably the most important change to an installation s capacity will be an upgrade to a faster CPU Modeling the replacement of a uni processor with another uni processor is fairly straight forward The parameters affected by this kind of change will be CICS CPU service rate CPUR and utilization of the CPU by tasks running higher priority than CICS UH Changes in these parame ters should be scaled to the ratio of internal speed of the 2 CPUs These numbers are published by IBM as tables of MIPS Ma chine instru
29. alties Inc Exclusive Distributors Axios Products Inc November 1 1999 E FACTORS BEST WORST OPTIONS MIPS RETURN SETUP EXIT te the slope of response times for varying transaction rates A B E D E F G H I Model Spreadsh Spreadsheet Version MODELCIC v1L5 0 The following short cuts are available ALT M This Menu ALT B Best Case Worst Case ALT R Response Time ALT O Option Comparison ALT F Sensitivity Factors ALT C CPU MIPS Conversion Use CTRL x in the Windows version of Lotus O CO JU WN p Model Input Parameters CPUR CPU Service Time UP Page Busy UH Super Overhead TP Page Service Tim UD Data Busy PFR Page Fault Rate TD Data Service Time TOR I O Rate XR Arrival Rate TITLE Current System Delta of Arrival Rate Ore TP UP TD UD IOR UH PFR 0 035 0 150 0 035 0 300 5 000 0 200 2 2 o ITAU E WNEF OC to 20 23 Nov 93 03 30 PM igure S M ODLCICS Template Main Panel CHANGE RSP_XRATE SLOPE_XRATE SAVE_GRAPH PRINT RETURN Change model input parameters for current system Press CALC ALT R for menu A B E D F G H I Model Spreadsh Spreadsheet Version MODELCIC v1L5 0 The following short cuts are available ALT M This Menu ALT Best Case Worst Case ALT R Response Time ALT Option Comparison ALT F Sensitivity Factors ALT CPU MIPS Conversion Use CTRL x in the Windows version of Lotus Dv oN
30. arting transaction rate The DXR value will be used to increment the transaction rate e Change the cell to the right of the label TITLE to describe this modeling run It will appear as a title in the graph e Press the S83 key to recalculate the de pendent cells This causes several model values to be recalculated First the transaction rate column is recalculated starting with the XR value and incre mented by DXR Next the response time values are recalculated based on the selected model input parameters and the calculated transaction rates To the right of the response time column is the slope of response time versus transaction rate These cells are also recalculated Final ly the sensitivity factors for each model input parameter are recalculated These values are stored in a different part of the spreadsheet This data can be viewed by and then area press the Sa key To display a graph of es time then press ME to select Graph Using the Slope and Factors Options e Ifyou select either SLOPE or FACTORS and then the CHANGE option you will be positioned over the same input area used by the RESPONSE option These three options use the same set of model input values Selecting GRAPH for the SLOPE option will display a graph of both response time and slope of response time versus transaction rate This provides a visual perspective of how transaction rate affects response time and where the k
31. atistics will be includ ed if the user fields are defined as stated in the CICS Resource Definition Guide JCL SMFPC110 is a stand alone program PCFMT EXEC PGM SMFPC110 REGION 1720K STEPLIB DD DSN TSU01 MYLIB DISP SHR SMFIN DD DSN BACKUP SMFWKLY G0001V00 7 VOL SER 123456 DISP OLD HM UNIT TAPE ISMFLOG DD SYSOUT A ISMF110S DD DSN TSU01 CICSSU M PRN DISP CATLG 7 SPACE TRK 4 4 RLSE JISMF110SR DD DSN TSU01 CICSRSP PRN 7 DISP CATLG 7 SPACE TRK 4 4 RLSE ISMFCTL DD COLLECT CICS TRANS EXCEPT SIGN ON AND CMF TRANS x EXCLUDE CCMF CSSN Notes 1 STEPLIB Used if SMFPC110 is not in the Link List 2 SMFIN Input data set of SMF records 3 SMFLOG Statistics on the run KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc 4 SMF110S CICS summary file to transfer to a PC 5 SMF110SR CICS summary response time file to transfer to a PC 6 SMFCTL Control card data set Spreadsheets There are two spreadsheets available for the analysis of CICS performance data The Summary spreadsheet data is obtained from the file created in the SMF110S file and is illustrated in Figure 1 on page 5 The fol lowing information is provided in the SMF110S file for each hour DATE Specifies the date CICS was executing The format can be mm dd yy or Lotus 1 2 3 date format D1 D2 or D3 TIME Specifies the time of day CID Specifies the CICS ID Th
32. ctions on the queue reaches 100 When the number of transactions on the queue drops below 100 the simulator will generate a new transaction for this work load Workloads are defined as either CICS IMS DB2 ADABASE IDMS or any other transaction request workloads Workloads are specified as up to 8 characters Each Workload is assigned a priority 1 20 A priority of 1 is the highest or 1st on the dispatch queue If two workloads have the same priority then their requests for resourc es will be equal The path length is defined for each workload as follows PILOT CICS 37 KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc Path length 10 MIP rate CPU seconds Transactions or requests where MIP rate is the relative speed of the processor on which the CPU time was collected CPU Seconds Is the total CPU time in seconds TCB and SRB the workload being measured consumed during the period Example Assume a CICS region used 30 5 mins of TCB time 5 1 mins of SRB time and did 2 300 transactions in one hour The total time is 35 6 minutes If the current system on which the region executed has a MIP Rate of 5 then we multiply 35 6 5 000 000 178 000 000 and divide by the number of transactions 2 300 to get 773 913 which is the path length Executing The Model After all parameters are provided the simu lator is started by pressing the F6 function key when the workload def
33. ctions per second A ratio of 2 to 1 new CPU MIPS to old CPU MIPS indicates the new CPU is twice as fast as the old CPU This means we can reduce the base value of CPUR by a like factor For example if the new CPU is 3 times faster than the current CPU we can calculate the new CPUR as old CPUR divided by 3 The same logic can be applied to the UH parameter but in real life a faster CPU will probably be executing more work This means the UH value will probably be re duced but not as much as was CPUR A conservative approach would be to scale UH by the ratio of MIPS and then add back in 10 20 of its new value For example modeling UH on a CPU which is 3 times faster than the current CPU can be done as follows assuming the current UH 15 let the new UH 15 3 plus 20 of 05 06 28 PILOT User s Guide November 1 1999 When modeling a new CPU which is a multi processor a different approach is taken First consider that CICS is a single tasking system This means that CICS can only execute on one processor at a time The value of CPUR must be computed as if the speed of the new CPU is equal to the speed of a single processor not the combined speed of all the processors High priority utilization UH however can be spread over all the processors so we can divide this value by the number of processors As an example assume a base value of CPUR 24 and UH 18 Assume an up grade to a dyadic processor with an aggrega
34. ctors Pane MENU CHANGE GRAPH SAVE_GRAPH PRINT RETURN Change parameters for Best Worst Case Press CALC ALT B to return to menu AL AM AN AO AP AQ AR AS AT AU Best Worst Case Capacity Plan Title Sample Best Worst Case Plan CPUR TP UP TD UD IOR UH Best 0 100 0 035 0 150 0 035 0 150 5 000 0 200 Worst 0 230 0 035 0 150 0 035 0 150 5 000 0 200 x vs RSP B RSP W MAX RSP Nov 93 L216 2 0 Dec 93 Jan 94 Feb 94 Mar 94 Apr 94 May 94 un 94 ul 94 Aug 94 Sep 94 23 Nov 93 03 34 P J U WNNNNNNNNDN N OO AN AU FPWNF O 0 4 0 4 0 4 0 4 0 4 0 4 0 4 0 4 0 4 0 5 0 5 UU UY 0YU1O OO HB Roo NUNNNNNNNN NWN O OO O O OO OSLO Figure 8 MODLCICS Template Best Worst Case Panel 32 PILOT User s Guide KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc November 1 1999 BF13 MENU CHANGE GRAPH MOVE SAVE_GRAPH PRINT RETURN Change model input parameters for different configurations Press CALC ALT O BF BG BH BI BJ BK BL BM BN Comparison of Different Configurations Title Sample Comparison Graph CPUR TP UP TD UD 100 035 150 300 071 030 150 A 300 100 030 150 A 300 100 015 100 200 D Q am OPTION OPT 4 MAX RSP 4 0 RESPONS ES OPT 1 pS bo N o tJ H H 4 114 114 UNO JO A KROE MS O O O OOGO G OMOORPRAA UY BS oO OOOO ene o IA UU Us Es gt NHOOOJOO
35. e time PILOT MVS s workload analysis makes this task trivial Do not be discouraged if your first predictions are wrong Instead study why they are inaccu rate and adjust those factors into your next forecast analysis For example take notes on when new applications are implemented Did the application use as much resources as planned Did the user do twice as many transactions as anticipated These are usual ly the causes of inaccurate forecasts Adding A New CPU This is the most significant change to a computer configuration that will yield the greatest impact on capacity and performance In addition to the speed of the computer usually the amount of memory is also increased Sometimes additional channels and strings are added thus giving better I O response times for workloads Each of these will be discussed separately After the model is calibrated a new CPU can be chosen from the list of processors provided with the simulator or a processor can be added if the MIP speed is known There are different MIP speeds reported for the same machines from various sources The main difference in the reported speeds is the relative performance of the computer to be considered 1 MIP The table used by the simulator is compiled from each vendor and other public information sources The relative speed at which the model was cali brated should be the same as the machine being replaced in the model Now execute the simulator with the new
36. e first four bytes of the NETNAME are used TASK The number of transactions processed for this period AVG RSP The average response time for all trans actions AVG WAIT The average wait time for all transactions AVG CPU The average CPU time for all transactions November 1 1999 OSTOR The average amount of storage used by all transactions in 1024 byte units AMC The total number of access method calls for all transactions I O The Total amount of I O processing all transactions performed This includes access method calls journal puts synchpoints BMS In BMS Out Tempo rary Storage AUX count and Transient Data get and put counts CPU CICS The total percentage of the CPU used by the CICS region CPU APPL The percentage of the CPU used by the application programs running within the CICS region PAGE IN The page in rate pages per second for the CICS region PAGE OUT The page out rate pages per second for the CICS region COMPS The number of storage bytes compressed during the period PCP RSP The average time for loading a program during the period PCP The number of program loads during the period PILOT CICS 3 KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc DLI RSP The average response time for a DLI call if DLI is installed DLI The number of DLI calls during the period if DLI is installed DB2 RSP The average response time for a DB
37. e month period for both best and worst case input The intersection of maximum acceptable re sponse time and these two lines represents the point where your installation s capacity is exceeded The time of year corresponding to these two points can be defined as the best case and worst case limits of CICS capacity Another way to define these times is to expect response time to begin to degrade at the worst case limit and to certainly become unacceptable at the best case limit If a twelve month projection fails to exceed your capacity limit then double the interval by treating each monthly value as a two month step Identify Performance Bottlenecks Once the capacity limits for the installation have been defined different configuration changes can be put into the model to fix the problem increase the limits These changes are generally limited to increasing one or more computer resources These include adding real memory upgrading the DASD sub system and installing a faster CPU Before selecting changes try to determine the primary bottleneck affecting response time This can be done by using the Factors option to calculate sensitivity factors Select the Factors option and select the Change sub option from the user menu The input parameters should be entered to reflect the model input values at the point of unaccept 26 PILOT User s Guide November 1 1999 able performance These values can be taken from the point where
38. e one or more of the 9 parameters to alter the model Some of the options will allow the user to select a starting CICS transaction rate XR and a delta transaction rate DXR This will allow the model to vary the transaction rate evenly beginning with XR and incrementing by DXR to pro vide a smooth curve Spreadsheet Features Options The MODLCICS spreadsheet includes 5 user options A list of these E can be displayed by pressing the Ha um the same time Please note that the Lotus 1 values are changed If the Cale indicator is displayed on the bottom line of the display then the spreadsheet needs to be recalculated Press Calc to do this If you are in a menu key must be used when the to get to a READY prompt November 1 1999 A sample of the template s main panel can be seen in Figure 5 on page 29 The 5 options are RESPONSE Calculates the average CICS response times for a range of CICS transaction rates A sample of the template s Re sponse panel and menu can be seen in Figure 6 on page 29 SLOPE Calculates the slope of response time versus transaction rate for a range of transaction rates These values can be used to estimate the breaking point out of capacity condition of the response time transaction rate curve A sample of the template s Response Slope panel and menu can be seen in Figure 6 on page 29 FACTORS Calculates sensitivity factors for each model input parameter The se
39. eation of a baseline model using RMF SMF and TMON data as input into the module gener ator program SIMBUILD Each file type must be separate The RMF records must be read first so that the program SIMBUILD can build a table in storage and match the CICS data to the RMF control performance groups Ensure that the time period of the input data from RMF matches the CICS monitor data Please note that RMF records are handled by the RMF start interval time stamp A separate step using PILOT SMF prior to model generation is recommended to select the correct RMF intervals using the TIME M parameter Also the multiple output function can be used to split the RMF and SMF 110 records if needed BUILD JOB ISIMCICS EXEC PGM SIMBUILD REGION 1500K ISTEPLIB DD DSN PILOT V1M5 LOADLIB DISP SHR SMFIN DD DSN RMF DATA DISP SHR SMF 70 75 DD DSN CMF DATA DISP SHR SMF 110 November 1 1999 MONIN DD DSN TMON DATA DISP SHR TMON USED PARMLIB DD DSN SYS1 PARMLIB DISP SHR SIMOUT DD DSN TSU MODEL FILE DISP CATLG SPACE TRK 1 UNIT SYSDA ISIMLOG DD SYSOUT ISIMCTL DD x Create baseline model for a CICS IMS environment x CPU MODEL 4381 2 WKLD WL MVSPGN 0 5 9 PRTY 1 WKLD WL PRODCICS 07 PRTY 2 APPLID N ODE25 DISKS VSAM1 INV PAY WKLD WL PRODIMS 08 PRTY 2 DISKS IM S DISKS IMSRES IMSQ WKLD WL IMSREG1 09 PRTY 3 IMSDB DISKS IMS0 INV MFG WKLD WL IMSREG2 10 PRTY 3 IMSD
40. ed is higher than 95 percent of the transactions processed for this period 75 RSP The 75th percentile of the transaction response time for this transaction code 50 RSP The 50th percentile of the transaction response time for this transaction code 25 RSP The 25th percentile of the transaction response time for this transaction code MIN RSP KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc The minimum transaction response time for this transaction code MAX RSP The maximum transaction response time for this transaction code 95 WT The file control wait time of the 95th percentile value as described above 50 WT The file control wait time of the 50th percentile value as described above MIN WT The minimum file control wait time for this transaction code MAX WT The maximum file control wait time for this transaction code CPU The average amount of CPU time used by the transaction STOR The average amount of storage the trans action used in 1024 byte units AVG I O The average amount of I O processing the transaction performed This includes access method calls journal puts synchpoints BMS In BMS Out Tempo rary Storage AUX count and Transient Data get and put counts PAGEIN The average number of page faults for this transaction CID November 1 1999 Specifies the CICS ID The first four bytes of the NETNAME are used PILOT CICS 5 KLM Technical
41. ee oe ta tato oe eee y Sloss 2 CXS Ite ss a eta Me Ma jeer tata a te eg whe ah 2 10 exclude yp rl eet aan eben la nth DEAR E Saal Sate be tie cl Meal rade Akh deat tomy ate 1 2 10 CNS cash ena o a Gl 35 ral ds so es Gee GS han hy de ou e a a a n a Da 5 7 14 29 TUNG ARANSIOR A oer ete ota we ene NN 33 adate NI BR See AM BE rn a He ate ir bs 2 10 PA A a es a ee Se aC en en ae ee eee 20 24 29 31 BRADING Ga tia a ay at ad ec 20 PEU A aoe ee Sn ha weap a ig e ata e 1 2 10 18 35 42 1 o vs estat A o o Yak Gy bay paro et RR q 41 43 POE ST SAA nie dios Er 5 7 14 1 g sg taste oe dod GT Ia wy oo Slush wed or AE GS he ee wate de ae 2 10 22 25 37 39 interval o ene aina ee Ce pb aa a E AR aa 9 23 24 34 43 45 47 WC OS Fa Sry Dr Th SG eB Soo Spe RR do To A 2 TA E AS 2 10 43 A E E EE Spe teed 42 KeyWords tie a ts e io ia e eae eee 1 10 41 AN NN 2 4 7 10 12 14 17 19 20 29 30 methodology a io o Sh ees 17 20 22 33 36 37 40 MPSS mal CE elas Sheil alae ee te Co a to dy dt 26 29 31 34 41 42 45 47 model A A wend Sir aa ete whan O ahh 17 19 27 29 31 33 34 36 43 45 47 PILOT CICS 51 KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc November 1 1999 model ENETALONS A tad 17 19 33 36 38 41 45 47 Model a ts e Gees 29 30 MIONE ts tad dto 9 10 ca o A A RN oe Abas 43 cup NA o a 1 9 pagesi misaa A ieee aS 3 5 11 13 18 21 23 26 29 30 33 36 39 40 45 47 48 PM a as aa
42. eed per processor not the combined speed of multiple processors OF CPUs The number of processors for the CPU model OF WLs The number of workloads to model Only workloads with higher priorities 36 PILOT User s Guide November 1 1999 need to be entered even though a lower priority workload may be the particular workload to be modeled Lower priority workloads below the workload of interest need not be modeled PAGE PACK SERVICE TIME The average service time in seconds for all paging devices for the system on which the workloads execute PAGE PACK BUSY The average percent busy of all paging devices for the system on which the workloads execute Function Key Definitions F1 HELP Help screens with brief descriptions of fields F2 NEXT Page forward to the workload definition screen F3 SAVE Save the current values in a configuration file F4 RETRIEVE Retrieve a previously saved configuration file F5 CPUTABLE Pick a new CPU to run with the model F10 END PGM End execution and return to DOS KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc Workload Definition Screen The next screen presented to you is the Workload Definition Screen A sample of this screen can be found in Appendix A in Figure 12 on page 46 The fields of this screen are defined here WORKLOAD The name assigned to each workload This name can be typed by you created by the SIMBUI
43. een presented to you is the Parameter Selection Screen A sample of this screen can be found in Appendix A in Figure 11 on page 45 The fields of this screen are defined here TITLE FOR THIS RUN This is a 40 character field to describe the model simulation TOTAL RUN TIME sec The total wall time to be simulated The default is 180 seconds This should be PILOT CICS 35 KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc enough time to model most environ ments More time may be required to simulate a low volume small transaction rate with CPU s of MIP rates 3 or less NOTE the program is simulating a CPU wall second this may in fact take more or less actual time to generate the results TIME INTERVAL SEC The time slice chosen to take a snapshot to simulate the given environment The smaller the value the more accurate the results The recommended values are CPU SPEED VALUE gt 3 MIPS but lt 7 MIPS 0001 Greater than 7 MIPS 0005 OF INTERVALS REPORT The number of time slices before a status line is written to the terminal This is used to monitor the execution of the simulator The default is 200 CPU DESCRIPTION The CPU model name of the processor A table of models and speeds are provid ed This field can be edited for pro cessors not yet included in the table CPU SPEED mips The machine instructions per second of the processor chosen This field repre sents the sp
44. either best case or worst case projections cross the maximum acceptable response time line Press S83 recalculate the sensitivity factors for each model input parameter These values can be viewed by selecting the GRAPH option or to can displayed by pressing S83 GOTO type SF then press enter The sensitivity factors SF represent the percent increase in re sponse time caused by a 1 percent increase in each input parameter The higher the sensitivity factor the more sensitive is response time to that input parameter These values can be used to identify which computer resource is the dominant cause of poor performance For example a high value for the page fault rate SF would indi cate a real memory constraint High SF values for disk service times and disk uti lization parameters TD and UD would indicate a problem with the DASD subsys tem And a high value of SF for CPU service time CPUR would certainly indicate that response time is sensitive to the speed of the CPU You can use the model to examine the effects of relieving these bottlenecks For example if a real memory constraint is indicated reduce the page fault rate PFR parameter in the input field and calculate the new response time A sharp reduction in response time will verify the bottleneck caused by excessive paging Compare Configurations to Solve Ca pacity Problems Based on the calculation of sensitivity fac tors resources affecting perfor
45. erisk is omitted or incorrectly specified the step will be terminated with a completion code of 16 Example 1 2 3 4 5 123456789012345678901234567890123456 78901234567890 EXCLUDE CSSN TRANO5 NO STATS FOR THESE TRANS Note that the parameters start in column two and that NO STATS FOR THESE TRANS is a comment EXCLUDE Tran1 Tran2 Tran3 E Specifies a group of transactions that are to be excluded from processing Up to 50 transactions may be specified PILOT CICS 1 KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc INCLUDE Tranl Tran2 Tran3 I Specifies a group of transactions that are to be included for processing Up to 50 transactions may be specified IREGS Reg1 Reg2 Reg3 Rega Specifies a group of CICS regions VTAM ACB name is used to be included for processing Up to 50 regions may be specified EREGS Reg1 Reg2 Reg3 Reg4 Specifies a group of CICS regions VTAM ACB name is used to be excluded for processing Up to 50 re gions may be specified Parameters on the JCL EXEC Card GDATE Specifies that the date is to be printed in MM DD YY format instead of Lotus 1 2 3 D1 format COMREGS Specifies that all input data found is to be processed as one output file combining all regions found SEPREGS Specifies that data is divided by regions found on the input file 2 PILOT User s Guide November 1 1999 DLI Specifies that DLI st
46. f CPU cy cles This is very important when you are examining many alternatives and playing the what if game and e Second a queuing model produces a non linear result of response time versus resource utilization This is in fact how real on line systems be have Models based on linear ap proximations are generally only accurate for a narrow range of re source utilization MODLCICS provides 5 user options A list of these options is displayed when the user menu is invoked This is done by pressing the Alt and M keys at the same time To gether these options provide the tools for automating a capacity planning methodology for CICS Model Parameters The nine input parameters for the MODLCICS spreadsheet are CPUR The average CPU service time for CICS transactions This value is expressed in PILOT CICS 19 KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc seconds and can be calculated by dividing the total CPU seconds used by CICS by the total number of completed transactions These measurements should be made for a reasonable period of time 1 hour is recommended This value is obtained from the PILOT CICS summary spreadsheet The average service time for the paging packs Also expressed as seconds this value is obtained from the PILOT MVS Page data set spreadsheet or can be read from an RMF disk subsystem report Since there will probably be more than one page pack defined this
47. he number of program loads during the period DSA_HWM The highest amount of storage allocated from dynamic storage area during the period In 1024 byte units DLI RSP The average response time for a DLI call if DLI is installed DLI The number of DLI calls during the period if DLI is installed November 1 1999 DB2 RSP The average response time for a DB2 call if DB2 is installed DB2 The number of DB2 calls during the period if DB2 is installed UDB_RSP The average response time for a user database IDMS ADABASE etc call UDB The number of user database IDMS ADABASE etc calls during the period INTVL The number of seconds in the period The Response spreadsheet data is obtained from the file created in the SMFMONSR file and is illustrated in Figure 4 on page 14 The following information is provided in the SMFMONSR file for each transaction processed DATE Specifies the date the transaction was initiated by CICS The format can be mm dd yy or Lotus 1 2 3 date format D1 D2 or D3 TIME Specifies the time of day the transaction was initiated by CICS in hhmm format TRAN The transaction name T Transaction types as follows A Auto initiated transaction P Printer transaction PILOT CICS 13 KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc T Terminal initiated transaction Z Conversational transaction R Inter region transaction I Inter system
48. ing modeled and other workloads that use a significant amount of resources We begin by developing profiles of the current work load resource usage This will help you determine the resources required by these workloads in the future The model genera tor provided in PILOT CICS will build a profile of the workloads you have defined SIMBUILD program In order to profile these workloads various sources of data must be analyzed Using PILOT MVS and PILOT CICS together will provide the data needed for this process If PILOT MVS was KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc not purchased this data can be obtained from RMF or other tools that provide similar reporting capabilities Creating the Base Line Model The second step in the methodology is the creation of a base line model This means defining the input values such that the simu lator will act like the current environment Although the input to the simulator contains average values such as average path length average arrival rate etc the simulator uses a randomizing technique so that different values are chosen for each sample or time slot Over time the average values for these parameters will approach the values you chose as your input This technique is required to simulate real systems with ran dom patterns of demand and service requests Identify Peak Periods A time frame to model must be chosen Choosing a time frame of 24
49. inition screen is displayed Once the model simulator initializes and begins execution a series of messages is displayed Each message pro vides information about the current snapshot of the model execution An intermediate report can be generated by pressing Maa This is recommended in order to verify the model parameters are correct For example a number which should have been entered as 025 and was incorrectly entered as 250 will have a significant effect on the results obtained 38 PILOT User s Guide November 1 1999 Image TO FREEZE THE DISPLAY PRESS foe S or mm nage vai After the intermediate report is reviewed execution is resumed by pressing the space bar The intermediate report and the final report have the same format A sample of the intermediate or final report is provided in Appendix A on page 47 Methodology The following methodology can be used to create a capacity plan for an online environ ment The purpose of this function is to predict future hardware requirements due to the increase in workload resource consump tion This is only one component of a complete capacity management discipline Creating models are useful when the analyst knows the questions A complete approach to capacity management and performance management are discussed in the techniques section of this manual Workload Characterization First one needs to understand the present resource consumption of the workload be
50. is included SMFCTL Specifies a control card file to indicate the regions and transactions to be processed If this file is omitted all regions and transactions are processed MONPCLMK and SMFPCLMK have at least two output files PILOT CICS 9 KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc This file is always required This file provides information on the data processed SMFMONS This optional file provides summary infor mation for each region processed When this DD stastement is omitted a file for downloading is not created SMFMONSR This optional file provides summary in formation for each transaction processed When this DD statement is omitted a file for downloading is not created Do not run these programs as user exits of READSMF when processing non SMF data Control Cards and Parameters There is one keyword with three operands that may be specified in the SIMCTL file The REGION keyword must start between columns 2 and 71 and be followed by a blank Operands for these control cards must be separated by a comma The REGION control card is needed to reduce output The format of the control card follows REGION INCL t1 t2 t3 IREG r1 r2 r3 EREG r1 r2 r3 The statement may be continued on the next card No special continuation character is required 10 PILOT User s Guide November 1 1999 KLM Technical Specialties Inc Exclusive Distributors Axi
51. mance can be determined and consequently a solution may be found to the problem It is possible to examine configuration changes to extend CICS capacity Selecting Options allows you KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc to model up to four configuration changes at the same time The graph produced for this option allows you to compare the relative benefits of each option The intersection of maximum acceptable response time and each response time transaction rate curve indicates the capacity limits for each option The transaction rate below each intersection is the maximum CICS load supported by that option The percent increase in the transac tion rate limit for option two over option one represents the percent increase in capacity of option two over option one If you divide this number by the cost of option two you have the percent increase in capacity per dollar spent Comparing these numbers will show you the relative cost of each option Assuming more than one option meets your objective for extending CICS capacity the most cost effective option can be chosen Choosing the Model Input Param eters The value of the CICS model is its ability to predict response time for a variety of config uration changes But before the model can be used you must be able to translate these changes into the input parameters understood by the model This section will offer advice on how that is done
52. nd should include at least five days of data This will ensure that you are not measuring an unusual day or spike condition If the CICS system makes use of conversational mode transactions then the tracking tool must report transactions and response time accordingly This means that a long conver sational mode transaction must be reported as several shorter transactions These values should represent internal response time and should not include think time or terminal network delay If the model s predicted response time falls within 20 of measured response time the model can be considered calibrated An error greater than 20 can be caused by several factors First ensure that the input KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc parameters and CICS response time were measured correctly A misplaced decimal point or wrong use of units i e transactions per minute not per second will certainly invalidate the model Another cause may be an internal bottleneck within CICS Internal bottlenecks represent conditions which degrade CICS performance despite the availability of CPU cycles memory and DASD resources Some examples are improperly set values of MAXTASK VSAM buffers and IMS DATABASE strings These represent artificial constraints to performance and cannot be accounted for in the model Ina sense these bottlenecks represent tuning problems and must be separated from capac ity planni
53. nee of the curve appears For example a value of slope equal to 2 means that the average CICS response time is increasing twice as fast as transaction rate A rule of thumb is to call this point the knee of the curve See the methodology discussion on page 22 for additional advice on how to use this data The values for slope are displayed in a column to the right of the response times The sensitivity factor values are displayed in a different part of the spreadsheet as described above PILOT CICS 23 to select the Remon option GOTO key then SF To return to the input then KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc Using Best Worst and Options Selecting BEST WORST or OPTIONS and then CHANGE will position the cursor over the input values for these options BEST WORST case input values consist of two sets of model input parameters Any of the model parame ters can be changed to indicate a best and worst case set of conditions Values can also be selected for transaction rates corresponding to twelve month s growth SES will recalculate the response times for each of the twelve months for both best and worst case values By changing the value for MAX RSP maximum acceptable response time the capacity limitations can be forecasted Displaying a graph of BEST WORST will show the average response time for both best and worst case values plotted along with MAX RSP The inter
54. ng issues Calculate the Maximum Acceptable Re sponse Time Before creating a capacity plan one must define and quantify capacity exceeded This can be done by calculating the maxi mum acceptable response time for your installation A good rule of thumb is to locate the knee of the curve of response time versus transaction rate If the previous rule of thumb is used this occurs where the slope of response time equals 2 As noted earlier this represents that point on the response time transaction rate curve where response time is growing twice as fast as transaction rate Since this is an exponential curve the expected response times become unstable from this point forward Use the SLOPE option to calculate response times and slope of response time for varying transaction rates The response time value which corresponds to a slope of 2 is the value to look for In plain language this value November 1 1999 represents the maximum acceptable response time for your installation This value is installation dependent If the base values of the model input parameters are changed the value of response time which corresponds to a slope of 2 will also change Create a Capacity Plan Based On Ex pected Transaction Rate Growth A capacity plan can be created which will identify the capacity limits of your current installation This is done by projecting transaction rate growth over a twelve month interval Start with a base value
55. nsitivity factor for each parameter represents the percent change in response time caused by a 1 percent increase in the input parameter For example a sensitivity factor of 5 for page fault rate PFR means that a 1 percent increase in page fault rate will cause a 5 percent increase in response time A sample of the tem plate s Factors panel and menu can be seen in Figure 7 on page 30 BEST WORST Calculates a best worst case capacity plan based on a 12 month projection of transaction rates Best worst case condi tions are expressed as 2 sets of model parameters Transaction rates are pro jected over 12 months and the model PILOT CICS 21 KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc November 1 1999 calculates both best and worse case response times You can also enter a maximum allowable response time value This value will be graphed along with the best worse case response times to show where the capacity of the CICS system is exceeded A sample of the template s Best Worst Case panel and menu can be seen in Figure 8 on page 30 Additional information on the use of this option can be found in methodology dis cussion on page 22 OPTIONS Calculates 1 to 4 sets of response times for a range of transaction rates Each set of response times is derived from a different set of model parameters Up to 4 configurations may be specified The graphs of response times versus transac tion rate
56. on to Modeling PILOT CICS provides two models The first MODLCICS is an analytic queuing model to calculate average response times based upon 9 input parameters The second SIMCICS is a model used to simulate a multi tasking operating system by defining transaction request oriented workloads MODLLCICS executes under Lotus 1 2 3 and can be used to model simple CICS environ ments More complex environments require more knowledge about the expected results and the input factors can be difficult to estimate SIMCICS has two components a mainframe model generator and the actual simulator that executes on the PC The simulator can model more complex environments MODLCICS Spreadsheet The spreadsheet MODLCICS provides an automated tool for capacity planning and performance analysis of the CICS environ ment This program is an analytic queuing model which calculates average CICS re sponse times based on 9 input parameters These parameters represent average utiliza tion factors affecting memory the DASD subsystem and CPU use The model was originally introduced in an article appearing in the December 1982 issue of the IBM Systems Journal November 1 1999 Modeling The value of using an analytic queuing model for capacity planning is twofold e First an analytic model consists of equations which yield a single result i e response time when executed Analytic models can be run over and over with a minimal use o
57. onse Time Templat PILOT CICS C 1986 1993 1 2 3 4 5 6 7 8 9 There are macros you can use to return to the menus Press T A to bring up the query menu G to bring up the graphs menu M to bring up the main menu this screen P to bring up the print menu Use CTRL in the Windows version of Lotus A A A A o ITAU E WU HO to 20 23 Nov 93 03 20 PM Figure 2 Response Template for CMF PILOT CICS 7 KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc November 1 1999 8 PILOT User s Guide KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc November 1 1999 PILOT CICS for The Monitor Introduction SMFLOG PILOT CICS allows users to format The Monitor detail records for capacity planning and performance tuning on the system and transaction level Two versions of the program are provided The program MONPCLMK executes as a stand alone program to process the records created by The Monitor Version 8 The program SMFPCLMK executes as a stand alone pro gram to process records created by Landmark s database utility program DBUTIL and supports all earlier versions of The Monitor Both programs read The Monitor s detail and system interval re cords MONPCLMK and SMFPCLMK have two input files LMKIN Specifies the input file that contains The Monitor records SMFIN data files are not processed when this DD statement
58. or a DLI call if DLI is installed DLI The number of DLI calls during the period if DLI is installed DB2 RSP The average response time for a DB2 call if DB2 is installed DB2 The number of DB2 calls during the period if DB2 is installed UDB_RSP The average response time for a user database IDMS ADABASE etc call KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc November 1 1999 UDB The number of user database IDMS ADABASE etc calls during the period PGIN The average number of page faults for this transaction PGOUT The average number of page outs for this transaction CID CICS system identifier specified in the DFHSIT 1 7 or DFHTCT 2 1 SYSIDNT parameter PILOT CICS 15 KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc November 1 1999 Al W9 MENU IMPORT GRAPHS PRINT RETURN SAVE SETUP Load a PRN file for a CICS Summary Report A B E D E F G H T J Spreadsheet Version SUMCICS v1L5 0 1 2 3 4 PILOT CICS Summary Template 5 6 7 AXIOS PRODUCTS INC PILOT CICS C 1993 TMV8 KLM Technical Specialti CICS REGIONS TS04 DATE TIME CID TASKS AVG_RSP AVG_WAIT AVG_CPU OSTOR DSA AMC 07 14 93 1900 TS04 T29 0 089 0 038 0 0124 678 11268 1 07 14 93 2000 TS04 184 055 022 0 0055 678 11268 109 1 07 14 93 2100 TS04 300 093 023 0092 678 11268 1334 13 07 14 93 2200 TS04 403 295 058 9881 67
59. os Products Inc Example 1 2 3 4 123456789012345678901234568901234568 9012345 REGION EXCLUDE CSSN TRANOS5 NO STATS EXCL TRN1 TRN2 TRN3 TRN4 Where TRNx represents a transaction code to be excluded from processing An may be used as a mask in the last three positions Up to 50 transac tions may be specified INCL TRN1 TRN2 TRN3 TRN4 Where TRNx represents a transaction code Specifies a group of transactions that are to be included for processing An can be used as a mask in the last three positions Up to 50 transactions may be specified The EXCL and INCL keywords can be specified using an as a mask for trans action names For example INCL TR TI will include all transactions that start with TR or TI only IREG CID1 CID2 CID3 CID4 Specifies a group of CICS regions to be included for processing Up to 30 CICS IDs maybe specified EREG CID1 CID2 CID3 CIDA Specifies a group of CICS regions to be excluded from processing Up to 50 CICS IDs maybe specified November 1 1999 Parameters on the JCL EXEC Card GDATE Specifies that the date be printed in mm dd yy format instead of Lotus 1 2 3 D1 format COMREGS Specifies that all input data found is to be processed as one output file combining all regions found SEPREGS Specifies that data is divided by regions found on the input file JCL PILOT CICS for The Monitor is a stand alone
60. ot supported by SIMBUILD Creating a Baseline Model This step is straight forward if all the data is available for the time frame being analyzed First the peak period s must be identified This can be done by using PILOT MVS and PILOT CICS to choose the correct time period and at least 5 days 1 week of data preferably 20 days 1 month Tracking the data and choosing the trend as the input into the simulator is of prime importance The model generator can generate most of the parameters for the simulator Information will have to be added for workloads not supported by PILOT MVS or PILOT CICS For example if ADABASE is used as a database manager the number of requests can be added into the simulator at execution time SIMBUILD Parameters November 1 1999 Model Generator SIMBUILD is a mainframe program to create model parameters for PILOT CICS SIMCICS and PILOT MVS SIMMVS programs The more information available as input for a workload the less manual input required There are three control card functions that may be specified in the SIMCTL file The CPU and WKLD keywords must start be tween columns 2 and 71 and be followed by a blank Operands for these control cards must be separated by a comma The CPU control is needed to calculate the fields derived from the SRM constants in MVS The format of the control cards follows CPU MODEL name SPEED sss ss PROCS n MIPS mm mm ICS nn where CPU
61. processor One if November 1 1999 the advantages of the simulator is that machines can be compared with the same MIP speed and a different number of pro cessors For example a 6 6 MIP uni pro cessor will yield better results than a 6 6 machine with two processors if only one big CICS region accounts for majority of the resource consumption This of course is only one factor memory and I O still need to be analyzed as well as cost Adding New Memory Modeling memory is one of the most difficult tasks because there are a number of factors that effect memory management in a processor These factors include e User storage fences via IPS parame ters Program working set sizes Multi programming levels Paging data set placement The number of Paging data sets OPT SRM parameters Each of the above mentioned parameters can cause bottlenecks in the system and may cause a system to appear to be out of capaci ty Usually this is the first sign of a machine s need to be tuned As mentioned earlier performance bottlenecks are a sepa rate issue and are not addressed in this section The simulator s parameters for modeling memory changes are the page fault rate demand for each workload the service time and the utilization of the paging subsys tem The approach is simple but effective results are obtained Each workload must be re viewed in the baseline model Predicting a paging rate is difficult but some basic as s
62. regions database managers external to CICS or other CICS terminal owning regions can be analyzed using this tool The value of using a simulator is that it will take the average arrival rate of the transac tions and produce random values above and below the average rate to behave more like a typical online environment Using a simulator will provide non linear results In the past this technique would use an enor mous amount of resources and time Today personal computers can be used to run a simulator obtaining good results The results of the simulator is a snapshot of the CPU utilization of each processor the response times for each transaction or request processed the number of transactions or requests processed and the queue length for each workload SIMCICS has two components a mainframe model generator and the actual simulator which executes on the PC Parameters November 1 1999 SIMCICS Simulator After the base line model is created on the mainframe it must be downloaded to the PC For further information on downloading refer to file transfer section in this guide Refer to the Methodology and Model generator sections of this guide for Base line models To start SIMCICS on the PC 1 Ensure that you are in the correct DOS directory 3 To execute the simulator SIMCICS type SIMCICS fag 4 Enter the first screen option C olor or M onochrome Parameter Selection Screen The first scr
63. s can be used to judge the rela tive improvement in one option over another Each option is expressed as a set of model input parameters Once again a maximum allowable response time can be input to show the capacity limits of each option A sample of the template s Options panel and menu can be seen in Figure 9 on page 31 Each option can be selected from the com mand menu using the first letter of the option i e F for FACTORS or by pressing the cursor keys to highlight the option and then pressing the MY key A description of each highlighted option will appear on the second line When an option is selected a sub menu will appear All of the sub menus are the same They provide 2 user options These are CHANGE and GRAPH 22 PILOT User s Guide CHANGE Alter one or more of the input parameters for the currently selected option Lotus 1 2 3 will automatically position the cursor at the first model input parameter for this option This will typically be the CPUR value At the same time the user menu will be erased from the command line This means the full facilities of Lotus 1 2 3 are available to change the input parameters To display the menu again you must press AEREA The spreadsheet has been initially set with the protection switch enabled This means that only the fields marked as unprotected can be changed These fields will show up as high intensity on a monochrome display and as white on a color display
64. section of the BEST WORST lines with the MAX RSP line represents the month when CICS capacity for each option is exceeded The input area for OPTIONS is similar to BEST WORST but includes four sets of input values These correspond to the four options that can be modeled i e different CPU more memory etc The values for transaction rate are calculated from a starting value of XR and incre mented by DXR Displaying a graph of OPTIONS shows the curves of response time versus trans action rate for each of the four options The intersection of each curve with MAX RSP can be defined as the capacity limit for that option The value of transaction 24 PILOT User s Guide November 1 1999 rate at the point of intersection can be defined as the maximum load that option can sustain with acceptable performance Methodology This section describes a step by step meth odology which can be used to create a ca pacity plan for CICS The steps that follow use the capabilities provided in the five user options which make up the MODLCICS spreadsheet Calibrating the Model Calibrating the model means testing the model s ability to predict the current envi ronment This is done by tracking model input parameters using one or more tools 1 e PILOT CICS PILOT MVS CICS PARS and RMP and using the RESPONSE option to compare predicted response times against measured values Measurements should be made for a peak hour of CICS use a
65. tand alone program to process Type 110 records SMFPC110 has two input files SMFIN This DD statement specifies the input file This file is required SMFCTL This DD statement specifies a control card file to control the transactions to be pro cessed If this file is omitted all transac tions are processed SMFPC110 has at least two output files SMFLOG This file provides information on the data processed The file is always required SMF110S This file is optional It provides summary information for CICS on a global level When this file is omitted a file for down loading data is not created SMF110SR This file is optional It provides summary information for each transaction processed containing response times When this file is omitted a file for downloading data is not created November 1 1999 PILOT CICS For CMF Control Cards and Parameters Keywords for SMFPC110 may be specified in free format control statements The keywords are separated from other parame ters with a space All other parameters are separated by commas Keywords and pa rameters may appear between columns one and seventy one inclusive The statement may be continued on the next card No special continuation character is required Comments may be specified on any control card by placing an asterisk in column one making the entire card a comment or by leaving at least one blank on any control card past column seventeen If the ast
66. te MIPS equal to 4 times the current CPU Model this upgrade by again changing the values of CPUR and UH Assuming that the memory size is not changing the page fault rate is ignored Since CPUR is only affected by the speed of one processor set the new CPUR old CPUR 2 24 2 12 By dividing the higher priority work over both CPUs we can expect the CPU running CICS to experience a UH equal to old UH 4 half as much work on a CPU which is twice as fast Once again to be conservative add back 20 percent of the new number UH 18 4 20 18 4 054 If these numbers are put in the model it will show CICS capacity is not dramatically improved This is caused by taking advan tage of one half of the multi processor To optimize CICS capacity on a multi processor you must run more than one copy of CICS A second copy of CICS can be modeled as if it was running independent of the first copy one copy per processor with equal priority Assuming that CICS can be divided evenly you may run twice the maximum transaction rate calculated for a single copy of CICS KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc November 1 1999 But this is probably an unrealistic assumption Dividing a production copy of CICS into 2 equal parts is just not possible Even with the facilities of MRO Multiple Region Option it is difficult to create 2 equal parts from one CICS system The effective maximum transaction rate
67. te toes 33 Function Key Definitions se us 435 5 RA A A as 34 PILOT CICS i KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc November 1 1999 Workload Dermition Sereeity is a ae ee Ee Patas 34 Excounne Mie Model dE Gis ale oo Sie id ds de 36 WI ETHO COTO yasen ans at Rede eee va eee de cack lene 36 Workload Characterization 0 0 0 eee 36 Creatine he Base Ein Model Aia 37 Id ntity Peak Period Su 2 p ita pd Gly eee etapas lee eed PS aca a 37 Tracking Dit is weer las La das wees al oes 37 Model Gencrators cs ads hoses Gey Sih cath Sb Sr ners Seg sale Gee aha ow bora 37 Canbraune the Modeloces aia ES Boh cece is 37 Forecasting Future Hardware Requirements o ooo ooo eee 38 Identify Resource Utilization by Business USag o o ooo coco ooo 38 Adding A New CPU 150 ado rro ra UE Ea sad 39 Adding New MEMO ss a es 39 Changing DASD Devices be ws Ao Bathe tido 40 Model Sener e AA Ae oh ba ee ah ce 41 Introduction een eE o io 41 Creatina a Baseline Model as aia add da 41 SIMBUIED Parametros 41 SIMBUILD ICE As e A A haa 43 APP A o nad dr sa die 45 e E AR ER 49 1i PILOT User s Guide KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc Introduction PILOT CICS allows you to format CICS CMF SMF Type 110 records for capacity planning and performance tuning on the global system and transaction levels The program SMFPC110 executes as a s
68. the response time as an internal response time This does not include individual think time or network and terminal delay time This is the same value reported by the simulator Reviewing response times and online system behavior during peak times makes the calibration step easier Model Generator The model generator can generate most of the parameters required as input to the simulator Information will have to be added for workloads not supported by PILOT MVS or PILOT CICS For example if ADABASE is used as a database manager the number of requests for ADABASE services can be added into the simulator at execution time Calibrating the Model PILOT CICS 39 KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc Calibrating the model means testing the model s ability to predict the current envi ronment This is done by comparing the base line values to the results from the simulator For example the CPU utilization of the workload generated by SIMCICS should match the CPU utilization as mea sured from PILOT CICS or PILOT MVS The total CPU usage from the simulator will match only if all workloads were grouped as OTHER in the workload for the model generator The response times should match the measured response times from PI LOT CICS The results should be within 10 If the results vary by more than 20 first ensure that the input parameters and CICS response time were measured correct ly A
69. umptions can be made Memory today is PILOT CICS 41 KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc bought in at least 4 megabyte increments which can have a significant impact on overall performance Aside from storage fences the higher a workload s dispatching priority the more likely the effect of real memory will be to reduce the paging rate This is obvious because the pages will be referenced more often and not be candidates to be paged out As the paging rate overhead is reduced the utilization and service times should improve Be conservative in reducing the paging rates and paging subsystem pa rameters in relationship to the amount of memory being added For example your main workload being modeled is experienc ing a page fault rate of 12 pages per second the paging subsystem has an average service time of 40 ms and 35 percent utilization 3350 s Further assume the machine is a 4381 2 with 12 megabytes of memory and you wish to add 4 megabytes total of 16 megabytes A conservative approach to this upgrade would be to reduce the page rate in half and the service time to some reasonable values such as 30 ms 20 percent utilization This is only one configuration other solutions can be easily modeled Model at least three possible scenarios with new memory to get good results Changing DASD Devices The simulator can be used to model DASD changes on workloads Modeling an upgrade of
70. value should represent a composite of all the active page packs The average utilization of the paging packs This value represents the fraction of time the pack is busy i e 30 for a pack which averages 30 busy The average service time for the data base packs Similar to TP above this value represents a composite service time for all the packs used by CICS for data base disk I O This data can be obtained from the PILOT MVS DASD spreadsheet which can selectively choose packs The average utilization of the data base packs Similar to UP above IOR The average I O rate for CICS transac tions This value represents the number 20 PILOT User s Guide November 1 1999 of real disk I Os executed on behalf of a CICS transaction This value can be calculated by dividing the total number of disk I Os issued by CICS to the data base packs by the total number of completed transactions The average utilization of the CPU by tasks which run at a higher dispatching priority than CICS Expressed as a fraction this value measures the CPU capacity which is not available to CICS Tasks running at a higher priority than CICS are typically VTAM JES and the operating system Unique installation conditions may require that another sub system e g IMS run at a higher prior ity than CICS This value can be esti mated by examining the PILOT MVS workload activity reports or RMF Tasks which run at a higher priority than CICS
71. workload during simulation CPUR The CPU time in seconds the workload used for this configuration NQ The total number of requests that had to be queued during the simulation Resp C The average internal response time used to process one request or transaction within the CPU Resp I The average response time used to process the I O for a request or transaction The last two fields added together yield the average total internal response time for a request or transaction within the workload 50 PILOT User s Guide KLM Technical Specialties Inc Exclusive Distributors Axios Products Inc November 1 1999 Index Accounts Ot teeth eke Sak aloha ra O Tp ake ei pl Ga 35 37 A apse tei ee de bt eae ye Saale Dag oe E 25 Capacity planning see s cd eee ue od ae hie aes AE ow A la 3 1 9 17 23 33 38 A E ee Basle steps RA CRIS Re oct Cercle q Me Bay he lay rans 31 CICS et ns e da 1 1 5 7 9 14 17 27 33 35 39 41 43 45 47 e AAA A Stk ce oe hw A S 1 2 4 5 7 41 43 CPUer i us 3 5 11 13 14 17 18 22 26 29 31 33 36 38 39 41 43 45 48 cpu modek a a A gh A ag ace ea Beg 34 41 GAT Yili RN A A o 8 23 A AA 17 18 23 25 38 40 dit iS eta ae dd dea ae 2 5 10 12 14 E A O E SR 3 4 11 12 22 37 LL pve te ar ha a ae ok taeda lo 1 4 9 12 42 43 ds ae sat aye EA eee E AA acl gta a ai Gok pe oe oe 33 35 43 dinear ea E O 2 3 12 13 A E E A SR EE EE S 33 34 EUR A A is ad 34 38 43 45 46 a li ds io dilo ai

PILOT/CICS User Manual

Contents

Download Pdf Manuals

Related Search

Related Contents