Previous Topic: 6.11 Track Memory Usage of Database WorkloadNext Topic: 6.13 Identify Transactions Used by CICS Users


6.12 Perform In-depth Statistical Analysis of Data


PROBLEM:

    At times it is necessary to understand behavior and
    characteristics in order to finalize how data is to be
    used.

    In this example, the CPU time consumed per hour by all
    transactions is analyzed.

    At most data centers, the large size of the CICCSUnn
    DETAIL file prohibits keeping the file, so this analysis
    must use the DAYS file.  Note that in this DAYS file,
    individual transaction data has been summarized by hour
    and user ID so that:

    1.  We must compare hourly CPU time totals. (The unit of
        the measurement used for comparison is not really
        important.)

    2.  We must remove the subdivision of measurements by
        USERID.


DATABASE FILES TO BE ACCESSED:

    &MCICD..CICCSU01
        Yesterday's CICS User Activity File. (Note that we
        assume the default account code structure is being
        used so that CICACT2 contains the transaction ID.)

SUGGESTED SAS FACILITIES TO BE USED FOR REPORTING:

    PROC UNIVARIATE to provide standard statistical measures.
SAS CODE EXAMPLE:

************************************************************;
* TASK CPU TIME STATISTICAL (UNIVARIATE) ANALYSIS          *;
************************************************************;

 %LET BY = CICACT2 HOUR;
 %LET BREAK = HOUR;
 PROC SORT DATA=&MCICD..CICCSU01 OUT=W1;
  BY &BY;
 DATA;
  SET W1;
  %CSUSUM;
 PROC UNIVARIATE PLOT;
  VAR CSUCPUTM;
 TITLE "STATISTICAL ANALYSIS OF TASK CPU TIME CONSUMPTION";


GENERATED OUTPUT REPORT:

    Figure 6-10 illustrates the output reports produced by
    the SAS code illustrated above.

    STATISTICAL ANALYSIS OF TASK CPU TIME USAGE                                                 12:29 Tuesday, September 7, yyyy   1     The UNIVARIATE Procedure   Variable:  CSUCPUTM  (Task CPU Time)                                 Moments                                                Basic Statistical Measures     N                        3175    Sum Weights               3175    --+---1----+---2----+---3----+---4----+---5----+   Mean               1.15556247    Sum Observations    3668.91085        Location                    Variability   Std Deviation      11.3452878    Variance            128.715554   Skewness           23.1243493    Kurtosis            693.293514    Mean     1.155562     Std Deviation           11.34529   Uncorrected SS     412782.825    Corrected SS         408543.17    Median   0.080755     Variance               128.71555   Coeff Variation    981.797872    Std Error Mean       0.2013463    Mode     0.000256     Range                  420.86884                                                                                              Interquartile Range      0.18348                Tests for Location: Mu0=0     Test           -Statistic-    -----p Value------     Student's t    t  5.739179    Pr > |t|    <.0001   Sign           M      1586    Pr >= |M|   <.0001   Signed Rank    S   2516189    Pr >= |S|   <.0001       Quantiles (Definition 5)     Quantile         Estimate     100% Max       420.868836   99%             14.595504   95%              1.832608   90%              0.814583   75% Q3           0.202585   50% Median       0.080755   25% Q1           0.019109   10%              0.003375   5%               0.001172   1%               0.000272   0% Min           0.000000                   Extreme Observations     ------Lowest------        -----Highest-----         Value      Obs           Value      Obs      0.000000     3034         153.424     2850    0.000000     2299         177.113     2854    0.000000      540         178.725     2874    0.000064      684         217.564     2855    0.000128      688         420.869      571

  STATISTICAL ANALYSIS OF TASK CPU TIME USAGE                                              12:29 Tuesday, September 7, 2010   2     The UNIVARIATE Procedure   Variable:  CSUCPUTM  (Task CPU Time)                               Histogram                          #  Boxplot       430.*                                                    1     *          .          .          .       290.          .          .          .          .*                                                    1     *          .          .*                                                    2     *       150.*                                                    1     *          .          .*                                                    4     *          .*                                                    3     *          .*                                                    1     *          .*                                                    3     *          .*                                                    9     *        10************************************************  3150  +-0--           ----+---+---+---+---+---+---+---+---+--           * may represent up to 66 counts                          Normal Probability Plot      430                                                  *         |         |         |         |         |         |      290         |         |         |         |                                                  *         |         |                                                  *      150                                                  *         |         |                                                  *         |                                                  *         |                                                  *         |                                                  *         |                                          ﯯ뻻**       10**************************************************          +---+---+---+---+---+---+---+---+---+---              -2        -1         0                  

Figure 6-10.  In-depth Statistical Analysis of Data