TOCPREVNEXTINDEX

Lund Performance Solutions


Quick Tour for MPE/iX Systems
In this tutorial, you will learn the following:
  • How to start Forecast
  • How to load a collection file
  • How to preview and validate a collection file
  • How to graph the information
  • How to specify workloads, CPUs, and gowth rates to be included in graphs
  • This section focuses on MPE data files. See "Quick Tour for HP-UX Systems." for HP-UX data files.

    Overview

    The Quick Tour in this chapter introduces first-time users to the features and functions of Forecast Capacity Planner. For users upgrading from earlier versions of the software, these tours provide a quick and easy way to get acquainted with the changes in the current version.


    Before beginning the tour, install Forecast Capacity Planner along with the appropriate sample files on your personal computer.

    Running Forecast Capacity Planner

    Run Forecast Capacity Planner on your personal computer as described in "Starting Forecast Capacity Planner".

    Loading a Collection File

    Forecast Capacity Planner creates a model of your system based on the data collection (*.col) files transferred from the host system. See "MPE/iX Host Data" for information on host data collection and data transfer.


    Your software includes a sample collection file (HP3000A1.col) for use with this tour.
    To load the collection file:
  • On the File menu, click Load/Validate model. The Open dialog box will appear (Figure 5.1).
  • In the Files of type list, click Collections (*col).
  • Click once on the collection file, HP3000A1.col, to select the file.


  • Figure 5.1 Open dialog box (showing *.col files)
  • Click Open to load the collection file.
  • Choosing a CPU Equivalent

    In some instances, the CPU naming conventions on the host will differ from those on the PC. When that happens, Forecast will generate the Choose CPU dialog.




    Figure 5.2 Choose CPU dialog box
    The Choose CPU dialog states which CPU type cannot be found by the Forecast application. In the example shown in Figure 5.2, the CPU in question is a 3000/957. To resolve the issue, select the equivalent CPU name from the pull-down list (for example, 3000/957) and click OK.

    Investigating Validation Messages

    As the loading process runs, you will receive a list of warnings and informational messages in the Validation Messages dialog (Figure 5.3). These messages provide information about workload groups that may violate one or more modeling algorithm assumptions. For information about these assumptions, see "Validation Failures".


    Figure 5.3 Validation Messages dialog box
    Normally, you would investigate the causes of these messages and determine what, if any, influence they have on your model. It may be necessary to repeat running the data collection process to collect another sample from the host system. Information about evaluating workload groups is provided in "Identifying and Characterizing Workload Groups".


    For this exercise, assume that the messages do not affect the validity of the model. Click OK to continue.

    Previewing the Unvalidated Baseline Model

    When the loading procedure is complete, the Main Program window will display. The name of the file will also appear at the top left corner of the Forecast window. The window contains the unvalidated data loaded from the host system. (The model is not validated until it is saved in the Model (*.mdl) format.) The Main Program window for an unvalidated model is divided into three panes:
  • The file manager pane.
  • The global unvalidated centers pane.
  • The global unvalidated workloads pane.
  • The content of each pane is described following the example shown in Figure 5.4.


    Figure 5.4 Main Program window panes (unvalidated model)
    The file manager pane
    The file manager pane is a representation of the unvalidated model in outline form.
  • To expand or collapse the outline, click the Plus (+) or Minus (-) icons (respectively) within the file manager pane.
  • To access detailed information, click the individual CPU, Disk, or Workload icons.
  • Clicking on Centers (or anything contained within) will cause the Workloads pane to be hidden. Clicking on Workloads (or anything contained within) will cause the Centers pane to be hidden. To re-display both panes, click on the file name (HP 3000 MPE/iX Tutorial A1).
  • To edit items on the file manager pane, double-click the associated icon.
  • The global unvalidated centers pane
    The global unvalidated centers pane displays the results for each CPU and disk drive in the model. To view detailed information about a single resource in the model, click the associated icon in the file manager pane.
    The global unvalidated workloads pane
    The global unvalidated workloads pane displays the results for each workload group included in the model. To view detailed information about a single workload group in the model, click the associated icon in the file manager pane.

    Validation Thresholds

    Normally, you would review the data in the Main Program window to determine if the model is a valid representation of the system’s baseline performance. For this exercise, assume that the model is valid.


    Acceptable validation thresholds for each system resource are listed in the following table.


    Resource
    Pane/Column
    Acceptable Validation Variance
    CPU utilization
     
    up to 15 percent
    Transaction/hour (throughput)
    Workloads/ Var: Tran/Hr
    up to 30 percent
    Response time
    Workloads/ Var: Resp. Time
    up to 30 percent
    Disk I/O/second
    Centers/ Var: IO/sec
    Variances for disk I/O/second and disk queue length are ignored because most real-world situations do not account for issues like database spreading.
    Disk queue length
    Centers/ Var: Avg. Queue Length


    If your numbers are out of range, review your workload definitions as described in "Identifying and Characterizing Workload Groups". Model numbers can be changed in the Edit Workload screen. To change your variances:
  • Select the workload group to revise.
  • Press Enter.
  • In the Edit Workload screen, make your changes.
  • Press OK.
  • Validating the Baseline Model

    Validate the baseline model you have created:
  • On the Forecast menu, click Validated Model.
  • In the forecast dialog box, click Yes to begin the validation process. Notice that the name of the document (top left corner of the window) changes from HP3000A1.col to HP3000A1.mdl.
  • On the File menu, click Save Model As....
  • In the Save As dialog box, click Save to save the model in the default location on your computer.
  • Graphing the Model

    The validated model created in the previous steps provides an accurate snapshot of your host system. The real power of Forecast Capacity Planner is its ability to predict future performance based on user-supplied what-if scenarios designed to illustrate the forecasting and graphing capabilities of the software.


    The examples in this section are based on the following assumptions:
  • Standard system performance tuning (defragging and balancing disk drives, repacking the database, etc.) will not significantly reduce CPU utilization or dramatically improve online response time, or reduce batch job completion time.
  • Workload definitions on the host system are unchanged.
  • Memory size and the type of disk drives are unchanged.
  • One CPU with Zero Growth Rate

    For this example, graph the model validated in the previous exercises and assume no change in utilization occurs for 24 months.
    To create the graph:
  • On the Forecast menu, click Forecast Options. The Forecast Options dialog box will appear, showing the Options tab.


  • Figure 5.5 Forecast Options dialog box: Options tab
  • In the Forecast Options dialog box, change the number in the Periods box to 24.
  • In the Growth Type list box, select Linear.
  • Verify the Period type is Month.
  • In the Workloads list box, verify that only the All workloads check box is selected.
  • Select the Adjust Results check box.
  • In the Graph Type list box, verify that CPU Util is selected.
  • In the Growth Option list box, click Override.
  • In the CPUs list box, verify that only the 3000/957 check box is selected (clear all other check boxes in the CPUs list box).


  • Figure 5.6 Forecast Options dialog box: Options tab with new settings
  • In the Forecast Options dialog box, click the Rates tab.


  • Figure 5.7 Forecast Options dialog box: Rates tab
  • Verify that Tiered rate and Growth rate are both set to 0 (zero).
  • Click OK to close the Forecast Options dialog box and save your changes.
  • On the Options menu, deselect the Prompt for titles option if it has been selected.
  • On the Forecast menu, click Graph Forecast. The forecast graph displays.


  • Figure 5.8 Baseline graph for one CPU with zero growth rate


    NOTE If the graph labels are overlapped, adjust the Font and Size parameters in the Graph Options dialog box from the Options menu (see "Fonts Options" and "Sizes Options").


    The upper line (XL Threshold) in the example graph (Figure 5.8) represents the CPU utilization threshold of 85 percent. 85 percent is the default threshold limit for MPE/iX systems and is generally considered the level at which response time and batch process completion rates become unacceptable. For information on changing threshold limits, see "Setting Thresholds".




    The lower line (All workload groups) in the graph represents overall CPU utilization for all workload groups identified in the model (3000/957). The modeled CPU utilization of 67.1 percent is approximately 18 percent below the threshold limit of 85 percent.

    Four CPUs with Zero Growth Rate

    In this example, compare the baseline forecast for your original model with three larger CPUs.
    To create the graph:
  • On the Forecast menu, click Forecast Options.
  • In the Forecast Options dialog box, select the following CPU check boxes in the CPUs list box:
  • 3000/957 (CPU selected in original model)
  • 3000/977
  • 3000/992-100
  • 3000/995-200
  • Click OK to close the Forecast Options dialog box and update your graph.




  • Figure 5.9 Baseline graph for four CPUs with zero growth rate
    The upper line (XL Threshold) in the graph represents the 85 percent CPU utilization threshold for MPE/iX. The other four lines represent the CPU utilization for each of the four CPUs, based on the workload groups identified in the original, baseline model.




    As expected, the larger CPUs show a lower utilization rate for the modeled workload groups. The 3000/995-200, for example, can handle the same workload groups with approximately 21 percent CPU utilization, compared with 67.1 percent CPU utilization on the 3000/957.

    Four CPUs with Five-Percent Growth Rate

    Repeat the forecast with a five-percent growth rate for all four CPUs.
    To create the graph:
  • On the Forecast menu, click Forecast Options.
  • In the Forecast Options dialog box, verify that the following CPU check boxes are selected in the CPUs list box:
  • 3000/957
  • 3000/977
  • 3000/992-100
  • 3000/995-200
  • Click the Rates tab in the Forecast Options dialog box.
  • Click in the Growth Rate text box and type 5.
  • Click OK to close the Forecast Options dialog box and update your graph.




  • Figure 5.10 Baseline graph for four CPUs with five-percent growth rate
    This example graph represents the modeled utilization rate for the 3000/957 relative to the rates for the larger CPUs at a growth rate of five percent over 24 months. Notice that the model does not simply add five-percent utilization to the rate for each month. The five-percent growth rate is a factor used to calculate the growth of each of the workload groups, subject to the modeled capacity of the system.




    For example, the plotted curve for the 3000/957 begins to flatten at the threshold limit of 85 percent. The amount of increase in CPU utilization is constrained by the model as the CPU approaches 100 percent utilization.




    Without remedial action, the 3000/957 will reach the threshold capacity in December 2000. Remedial action might include replacing the 3000/957 with a larger CPU, or improving system performance.

    Utilization by Workload Group for One CPU

    For the last example, graph your original model for the 3000/957 for each workload group, assuming a five-percent growth rate.
    To create the graph:
  • On the Forecast menu, click Graph Forecast to close the graph and return to the Main Program window.
  • Note the percent utilization for each workload group displayed in the global validated workloads pane.


  • Figure 5.11 Main Program window: global validated workloads pane
    The CPU utilization for each workload group should read:


  • NETBASE:
  • 3.0
  • NET SERVIC:
  • 4.5
  • UTIL:
  • 2.1
  • TELEMARKET
  • 5.8
  • CHICAGO
  • 13.2
  • ORDER ENTR
  • 19.1
  • JOBS
  • 19.4


  • On the Forecast menu, click Forecast Options.
  • In the Workloads list box, deselect the All Workloads check box. Select each of the seven workloads individually.
  • In the Forecast Options dialog box, deselect all of the CPUs except the 3000/957 in the CPUs list box.


  • Figure 5.12 Forecast Options dialog box, Options tab
  • Click OK to close the Forecast Options dialog box and save your changes.
  • On the Forecast menu, click Graph Forecast. The forecast graph displays.




  • Figure 5.13 Graph showing utilization by workload group for one CPU
    The utilization growth for each workload group is plotted based on the workload definition given in the data collection file and the constraints in the baseline model.


    As expected, the larger workload groups, JOBS and ORDER ENTR, contribute the most to the 85-percent system threshold.


    This type of forecasting can predict the impact of adding workload groups to an existing system. It is also useful in balancing workload groups across multiple systems.

    Individual Workload Group Growth

    What if, on top of your assessed 5% growth, you figured that one or more individual workload groups would increase as well? Forecast Capacity Planner allows you to change the growth rate of individual workload groups in order to see the final outcome. To change individual workload groups from the previous example:
  • Under the Forecast menu, click on Graph Forecast to deselect the option and close the graph.
  • Double-click on the row that contains the CHICAGO workload. This opens the Edit Workload dialog box.
  • Click on Growth Rate. This opens the Edit Growth Rate dialog box.


  • Figure 5.14 Edit Growth Rate (Individual Workload Group) dialog box
  • In the first row, change the rate to 5.00.
  • Change the months from 999 to 24.
  • Click OK.
  • Click OK again to exit the Edit Workload dialog box.
  • Under the Forecast menu, open Forecast Options.
  • On the Options tab, change the Growth Option to Add.
  • Click OK.
  • Under the Forecast menu, select Graph Forecast to open the graph.




  • Figure 5.15 Graph showing 5% individual growth for the CHICAGO workload group.
    As you can see by comparing this graph to the graph generated in the previous section, the CHICAGO workload group increases more noticeably. (ORDER ENTRY and JOBS show an increase only because they are stacked on top of CHICAGO.) At this rate of growth, CPU utilization will surpass the 85% mark by November of 2000.


    On the File menu, click Close. If you want to use this model again, save your sample file at the prompt.

    Lund Performance Solutions
    www.lund.com
    Voice: (541) 812-7600
    Fax: (541) 81207611
    info@lund.com
    TOCPREVNEXTINDEX