Capacity planners are frequently asked to forecast the potential impact of applications that are still in development. System designers and targeted end-users may supply estimated workload characteristics or transaction volumes based on comparisons to existing systems. You can develop early projections based upon estimates of this type using Univariate Model Forecasting. In this case study, an existing, inquiry-only online application is to be replaced by a system with both inquiry and update capabilities. It was estimated that during peak hour, inquiry transactions would account for 60% of the new system's transaction volume. It was further estimated that inquiry transactions would increase 15% with installation of the new system, but that overall transaction volume growth would not change. The purpose of this study is to provide an initial projection of the transaction volumes to be expected for the new application during the first three months following its installation. The estimates indicate that transaction growth should be modeled as a step function, with the step occurring in January 1999. Note that the magnitude of the step is estimated as a function of the transaction volume expected in December 1998. Since no data are available beyond June 1998, two models were developed. The first model was developed to project transaction volumes for the existing system through December 1998. Figure 7-23 shows the contents of the resource element file that contains monthly peak hour transaction volumes for the existing application. A linear model, which uses all 18 months of historical data was proposed. Figure 7-24 shows the control parameters that we used to produce the model. Figure 7-25 shows the resulting Residual Analysis Report.
CA MICS Capacity Planner OBS DATE MONTH CISTRANS SYSID ZONE YEAR 1 31JAN97 1 4043.0 IMS1 1 97 2 28FEB97 2 3890.0 IMS1 1 97 3 31MAR97 3 4037.0 IMS1 1 97 4 30APR97 4 4128.0 IMS1 1 97 5 31MAY97 5 4037.0 IMS1 1 97 6 30JUN97 6 4314.0 IMS1 1 97 7 31JUL97 7 4373.0 IMS1 1 97 8 31AUG97 8 4740.0 IMS1 1 97 9 30SEP97 9 4759.0 IMS1 1 97 10 31OCT97 10 4856.0 IMS1 1 97 11 30NOV97 11 4741.0 IMS1 1 97 12 31DEC98 12 5099.0 IMS1 1 98 13 31JAN98 1 5149.0 IMS1 1 98 14 28FEB98 2 4857.0 IMS1 1 98 15 31MAR98 3 5506.0 IMS1 1 98 16 30APR98 4 5196.0 IMS1 1 98 17 31MAY98 5 5194.0 IMS1 1 98 18 30JUN98 6 5477.0 IMS1 1 98
Figure 7-23. Contents of Transaction Volume Resource Element File /------------------------ Univariate Model Forecasting -----------------------\ |Command ===> | | | |Composing CA MICS Inquiry: UVMW2 - Univariate Modeling Inquiry | | | |Report title ===> Univariate Modeling Inquiry - CICS | | | |Selection criteria: | | Input file ===> CIS - CPU Analysis w/Composite SYSID/ZONE | | Start date ===> _________ (ddmonyyyy) | | SYSID ===> IMS1 | | Zone ===> 1________ | | No. weeks ===> ____ (1-9999) | | Save forecast ===> YES (YES/NO/AGE) | | Specify CAPAPU Values ===> N (Y/N) | |Forecast data element ===> CISTRANS | |Forecast length (weeks) ===> 6 (1-9999) | |Confidence limits (percent) ===> __ (70/90/95) | | | |Time element terms: Linear ===> YES (YES/NO) | | Quadratic ===> ___ (YES/NO) | | Cubic ===> ___ (YES/NO) | |Specify extended options ===> N (Y/N) | \------------------------------------------------------------------------------/ Figure 7-24. Control Parameters for Existing System Model
C A M I C S C A P A C I T Y P L A N N E R UNIVARIATE MODELING RESIDUAL ANALYSIS UNIVARIATE FORECAST OF: CISTRANS BASED ON TIME ELEMENTS: LINEAR 95% TRANSFORMED PREDICTED RESIDUAL -------------PLOT OF RESIDUALS------------- CONFIDENCE DATE CISTRANS CISTRANS CISTRANS CISTRANS -301 0.0 +301 LIMITS +/- ------- ---------- ------------ ---------- ---------- ------------------------------------------- ---------- 31JAN97 4043.00 4043.00 3889.39 153.61 | 0++++++++++ | 156.05 28FEB97 3890.00 3890.00 3983.42 -93.42 | ------0 | 142.91 31MAR97 4037.00 4037.00 4077.45 -40.45 | --0 | 130.33 30APR97 4128.00 4128.00 4171.49 -43.49 | --0 | 118.48 31MAY97 4037.00 4037.00 4265.52 -228.52 | ---------------0 | 107.62 30JUN97 4314.00 4314.00 4359.55 -45.55 | ---0 | 98.07 31JUL97 4373.00 4373.00 4453.58 -80.58 | -----0 | 90.25 31AUG97 4740.00 4740.00 4547.62 192.38 | 0++++++++++++ | 84.63 30SEP97 4759.00 4759.00 4641.65 117.35 | 0+++++++ | 81.68 31OCT97 4856.00 4856.00 4735.68 120.32 | 0+++++++ | 81.68 30NOV97 4741.00 4741.00 4829.72 -88.72 | -----0 | 84.63 31DEC97 5099.00 5099.00 4923.75 175.25 | 0+++++++++++ | 90.25 31JAN98 5149.00 5149.00 5017.78 131.22 | 0++++++++ | 98.07 28FEB98 4857.00 4857.00 5111.82 -254.82 | ----------------0 | 107.62 31MAR98 5506.00 5506.00 5205.85 300.15 | 0+++++++++++++++++++ | 118.48 30APR98 5196.00 5196.00 5299.88 -103.88 | ------0 | 130.33 31MAY98 5194.00 5194.00 5393.91 -199.91 | -------------0 | 142.91 30JUN98 5477.00 5477.00 5487.95 -10.95 | 0 | 156.05 31JUL98 . . 5581.98 . | 0 | 169.62 31AUG98 . . 5676.01 . | 0 | 183.53 30SEP98 . . 5770.05 . | 0 | 197.70 31OCT98 . . 5864.08 . | 0 | 212.08 30NOV98 . . 5958.11 . | 0 | 226.64 31DEC98 . . 6052.15 . | 0 | 241.33
Figure 7-25. Univariate Model of Existing Application
The second model was developed to project transaction volumes for the new application. Again, a linear model was proposed, with a user transformation defined to account for the anticipated step function. The magnitude of the step was calculated from the percentage estimates provided and the predicted transaction volume for December 1998 was obtained from the first model. Figure 7-26 shows the control parameters used to produce the model.
/------------------------ Univariate Model Forecasting -----------------------\ |Command ===> | | | |Composing CA MICS Inquiry: UVMW3 - Univariate Modeling Inquiry - CICS2 | | | |Report title ===> Univariate Modeling Inquiry | | | |Selection criteria: | | Input file ===> CIS - CPU Analysis w/Composite SYSID/ZONE | | Start date ===> _________ (ddmonyyyy) | | SYSID ===> IMS1 | | Zone ===> 1________ | | No. months ===> ____ (1-9999) | | Save forecast ===> YES (YES/NO/AGE) | | Specify CAPAPU Values ===> N (Y/N) | |Forecast data element ===> CISTRANS | |Forecast length (weeks) ===> 9 (1-9999) | |Confidence limits (percent) ===> __ (70/90/95) | | | |Time element terms: Linear ===> YES (YES/NO) | | Quadratic ===> ___ (YES/NO) | | Cubic ===> ___ (YES/NO) | |Specify extended options ===> Y (Y/N) | \------------------------------------------------------------------------------/ Figure 7-26. Forecasting for New Application Model (Screen 1 of 4) /--------------------------- Extended Options Menu ---------------------------\ |Command ===> | | | |Inquiry Step: Univariate Model Forecasting | | | |Print Reports ===> YES (YES/NO) | |Maximum R-Square for report ===> 0.55 (0.00-1.00) | |P-Value flag ===> 0.05 (0.000-1.000) | | | |Delete historical data observations ===> N (Y/N/R) | |Transform historical observations ===> N (Y/N/R) | | | \------------------------------------------------------------------------------/ Figure 7-26. Forecasting for New Application Model (Screen 2 of 4) /--------------------- Transform Historical Observations ---------------------\ |Command ===> | | | |Inquiry Step: Univariate Model Forecasting | | | |Smoothing technique ===> USER (GMA/LOG10/USER/NONE) | | | | | | | |Geometric moving average data smoothing parameters: | | Equation alpha value ===> ___ (.01 - .99) | | Initial value ===> ________ | | | | | | | |User defined data transformation SAS statements: | | Statements to transform historical data ===> Y (Y/N/R) | | Stmts. to reverse history transformations ===> N (Y/N/R) | | | | | | | | | | | \------------------------------------------------------------------------------/\ Figure 7-26. Forecasting for New Application Model (Screen 3 of 4) /-------------------------- SAS System Statements ----------------------------\ |Command ===> Scroll ===> PAGE | | | |SAS Statements for the ASGN_TRN Macro | |-------------------------------------------------------------------------------| |****** ***************************** TOP OF DATA ******************************| |==MSG> SPECIFY SAS STATEMENTS TO PERFORM A USER TRANSFORMATION. | |=NOTE= %MACRO ASGN_TRN; <=== PLEASE DO NOT PLACE LINES BEFORE THIS LINE. | |000001 _TRAN_V_=&VNAME+5569; | |=NOTE= %MEND ASGN_TRN; <=== PLEASE DO NOT PLACE LINES AFTER THIS LINE. | |****** **************************** BOTTOM OF DATA ****************************| | | \------------------------------------------------------------------------------/ Figure 7-26. Forecasting for New Application Model (Screen 4 of 4)
The special variable name _TRAN_V_ in the SAS statement on the Univariate Model Forecasting screen shown in Figure 7-26 refers to the transformed historical value of the dependent variable, CISTRANS, and the special variable name, &VNAME, refers to the value of CISTRANS prior to transformation. Section 7.5.4 describes these special variable names in more detail. Figure 7-27 shows the Residual Analysis Report for the model. Note that the residuals shown are identical to those in Figure 7-25. This is to be expected, since all historical data is transformed by adding the same constant value. Figure 7-28 shows a plot of the resulting composite forecast.
C A M I C S C A P A C I T Y P L A N N E R UNIVARIATE MODELING RESIDUAL ANALYSIS UNIVARIATE FORECAST OF: CISTRANS BASED ON TIME ELEMENTS: LINEAR 95% TRANSFORMED PREDICTED RESIDUAL -------------PLOT OF RESIDUALS------------- CONFIDENCE DATE CISTRANS CISTRANS CISTRANS CISTRANS -301 0.0 +301 LIMITS +/- ------- ---------- ------------ ---------- ---------- ------------------------------------------- ---------- 31JAN97 4043.00 9612.00 9458.39 153.61 | 0++++++++++ | 156.05 28FEB97 3890.00 9459.00 9552.42 -93.42 | ------0 | 142.91 31MAR97 4037.00 9606.00 9646.45 -40.45 | --0 | 130.33 30APR97 4128.00 9697.00 9740.49 -43.49 | --0 | 118.48 31MAY97 4037.00 9606.00 9834.52 -228.52 | ---------------0 | 107.62 30JUN97 4314.00 9883.00 9928.55 -45.55 | ---0 | 98.07 31JUL97 4373.00 9942.00 10022.58 -80.58 | -----0 | 90.25 31AUG97 4740.00 10309.00 10116.62 192.38 | 0++++++++++++ | 84.63 30SEP97 4759.00 10328.00 10210.65 117.35 | 0+++++++ | 81.68 31OCT97 4856.00 10425.00 10304.68 120.32 | 0+++++++ | 81.68 30NOV97 4741.00 10310.00 10398.72 -88.72 | -----0 | 84.63 31DEC97 5099.00 10668.00 10492.75 175.25 | 0+++++++++++ | 90.25 31JAN98 5149.00 10718.00 10586.78 131.22 | 0++++++++ | 98.07 28FEB98 4857.00 10426.00 10680.82 -254.82 | ----------------0 | 107.62 31MAR98 5506.00 11075.00 10774.85 300.15 | 0+++++++++++++++++++ | 118.48 30APR98 5196.00 10765.00 10868.88 -103.88 | ------0 | 130.33 31MAY98 5194.00 10763.00 10962.91 -199.91 | -------------0 | 142.91 30JUN98 5477.00 11046.00 11056.95 -10.95 | 0 | 156.05 31JUL98 . . 11150.98 . | 0 | 169.62 31AUG98 . . 11245.01 . | 0 | 183.53 30SEP98 . . 11339.05 . | 0 | 197.70 31OCT98 . . 11433.08 . | 0 | 212.08 30NOV98 . . 11527.11 . | 0 | 226.64 31DEC98 . . 11621.15 . | 0 | 241.33 31JAN99 . . 11715.18 . | 0 | 256.14 28FEB99 . . 11809.21 . | 0 | 271.05 31MAR99 . . 11903.24 | 0 | 286.03 * - DATA SMOOTHED USING USER DEFINED MACRO
Figure 7-27. Univariate Model of New Application
PEAK HOUR FORECAST OF ONLINE APPLICATION EXISTING APPLICATION FORECAST THROUGH DECEMBER, 1998 NEW APPLICATION FORECAST AS OF JANUARY, 1999 13000 + | ACTUAL VALUES SYMBOL IS * | FORECAST VALUES SYMBOL IS + | 12000 + + | + + | | 11000 + | | | 10000 + | | | 9000 + | | | 8000 + | CISTRANS | 7000 + | | | 6000 + + + | + + + | * * + | * * * 5000 + * | * * * * * | | * * * 4000 + * * * * | | | 3000 + | | | 2000 + -+-------+-------+-------+-------+-------+-------+-------+-------+-------+-------+-------+-------+-------+-------+-------+- JAN97 MAR97 MAY97 JUN97 AUG97 OCT97 DEC97 FEB98 APR98 JUN98 AUG98 OCT98 DEC98 FEB99 APR99 JUN99 DATE
Figure 7-28. Transaction Volume Forecast
|
Copyright © 2014 CA.
All rights reserved.
|
|