TOCPREVNEXTINDEX

Lund Performance Solutions

Performance Gallery Gold Quick Tour
In this tour, you will learn the following:
  • How to open data files.
  • How to open one or more charts.
  • How the charts relate to your system performance.
  • How to modify the charts to fit your needs.
  • How to save desktops for future use.
  • How to close the Performance Gallery Gold program.
    • Because the Windows NT platform is a relatively new addition to Performance Gallery Gold, we will be working with data files gathered from a Windows NT server. The steps, examples, and conclusions drawn from the data are readily applied to the MPE/iX and Unix platforms as well.

    Overview

    The Quick Tour in this chapter introduces first-time users to the features and functions of Performance Gallery Gold for Windows 2000/NT, MPE/iX, and Unix. For users upgrading from earlier versions of the software, these tours provide a quick and easy way to get acquainted with the changes provided in the current version, D.0x.
    Before beginning the tour, install Performance Gallery Gold along with the appropriate sample files on your personal computer as described in "Product Startup".

    The Charts

    We will open four charts during this tutorial which are designed to reveal different aspects of your system’s performance. No one single Performance Gallery Gold chart can tell you everything you need to know about your system performance. For instance, a high Processor Queue Length, at first glance, may indicate a CPU bottleneck and hence, a need for a faster processor. But perhaps the CPU Utilization never surpasses 50%. Obviously, the CPU is not being overworked. In this case, it is likely that several processes or services were started all at the same time. Purchasing a faster processor wouldn’t help in this situation; it would be a waste of your company’s money.
    CPU Utilization
    This chart displays how much of your Central Processor Unit (CPU) is being used. If this metric reaches 85%, the system will begin to slow down noticeably. If the CPU Utilization remains at 85% for extended periods of time, or spikes at certain times every working day, it can be an indication of a few things:
  • Too many users for the CPU. You may need to upgrade to more or faster processors.
  • You need to schedule batch and low priority processes or services to run at times (at night, for instance) when your users will not be using the system, thereby spreading the CPU usage around the clock.
  • You may need to add more memory.
    • For a more detailed look at CPU Utilization, see "CPU Utilization by Process".
    Logical Disk Response Time
    This chart displays how long it takes from the time a request for data (read or write) is made to the time the data is read from or written to the disk. Most requests can be satisfied within 11 ms (0.011 sec). If your response times are consistently above 0.025 seconds, it may indicate the need to:
  • Increase memory. If the data is found in memory, then there is no need to access the disk, and the response time will improve.
  • Upgrade to a faster disk. Faster disks transfer data faster.
  • Defragment your hard disk. If the data on your disk is spread throughout the disk, it will take longer to find the data you are interested in.
  • Invest in other hardware solutions. See "Logical Disk Response Time" for more information on these solutions.
    • File Cache Activity by Type
    This chart displays how many times per second a request for data was satisfied from the memory or cache, (also known as read/hits). If the read/hit ratio is high, that means data was successfully cached and accessed, thereby reducing the number of times the disk needed to be accessed to find data (disk I/O). Adding memory, defragmenting your system, or spreading out processes and services over a longer period of time may improve the File Cache read/hit ratio. See "File Cache Activity by Type" for more information.
    Processor Queue Length
    This chart shows how many processes (threads) are waiting for service in the NT dispatcher ready queue. Each processor should have a queue length of no more than 2 threads. A long queue length could indicate that:
  • Several processes are waking up and attempting to run all at the same time. These could be spread out to run at different times so as not to bunch up.
  • The CPU is overloaded and is causing a bottleneck of data.
    • MPE/iX and Unix
    Charts for Unix and MPE/iX systems that resemble those previously described are:
  • CPU Utilization
  • Disk Service Time by Drive
  • Memory Manager/Rd Hit%
  • Disk I/O Queue Length

    • Getting Started

    This section demonstrates how to start Performance Gallery Gold, how to open a data file, and how to open a chart. After these steps are completed, we can study and analyze the charts in order to assess the performance data.

    Starting Performance Gallery Gold

  • On your Start menu, click Programs.
  • Click on Performance Gallery Gold. Your program file may be nested in another folder, as shown in Figure 5.1.

    • Figure 5.1 Starting up Performance Gallery Gold

    Opening a New Data File

    Performance Gallery Gold creates charts of your system performance based on the SeNTry (*.smf) or *.pfg files transferred from the host system.
    Your software includes a sample data file for use with this tour.
    To load the data file:
  • On the File menu, click Open Data File. The Open dialog box displays (see Figure 5.2).
  • In the Files of type list, click SeNTry (*.smf).
  • Click once on the data file, NTSystem.199902201434.smf (the actual name may vary), to select the file. If you have your own SOS generated .pfg file, open it instead.

    • Figure 5.2 Open Data File dialog box
  • Click Open to open the data file. The name of the data file will be visible at the top of the Performance Gallery Gold screen.

    • Assigning an Alias

    Once you have opened a data file, you may be given the opportunity to rename the data file. This capability is most useful when data files from multiple systems have been loaded (see "System Alias Manager" for more information regarding Multiple Host Capability). For our purposes here, you may simply click Cancel in the System Alias Manager dialog box.

    Opening a New Chart

    Now that the data file has been opened, Performance Gallery Gold gives you options as to what information the chart displays. For this exercise, you will open several key performance charts. Make sure the Performance Gallery Gold window is maximized.
    To open the charts:
  • On the Chart menu, click Open Chart.

    • Figure 5.3 Chart menu with Open Chart selected
  • In the General tab of the Open Chart dialog box, select the chart titled CPU Utilization.

    • Figure 5.4 Open Chart dialog box with CPU Utilization selected
  • Click OK. The chart opens and the Edit and Window menus are added to the menu bar.
  • Repeat steps 1-3, opening the following charts:
  • Logical Disk Response Time
  • File Cache Activity by Type
  • Processor Queue Length
  • In the Window menu, click Tile Horizontal. The four open charts will tile so that each one has it’s own quadrant (see Figure 5.5).

    • Figure 5.5 All charts opened and tiled

    Analyzing the Charts

    In this section, we will look at the attributes of the four open charts. Once we see their default settings, we will modify parts of each graph to better acquaint you with all the options Performance Gallery Gold offers and where these options are found. Changes can be made to charts to focus in on certain time periods, to narrow the charts down to certain data elements, or simply to make them more readable and aesthetically pleasing.

    CPU Utilization Chart

    This chart displays how much of your Central Processor Unit (CPU) is being used. If this metric reaches 85%, the system will begin to slow down noticeably.
    Identify Chart Properties
    Graph Attributes
    Take a closer look at the chart. Maximize the CPU Utilization chart.
  • The left side of the graph, (the y axis) shows how much (what percent) of the CPU was being used during the sample intervals.
  • The x axis displays the length of the reporting period. In this case, the reporting period is from 14:34 to 16:07.

    • Figure 5.6 Example CPU Utilization graph from sample data
    System Performance Conclusions
    According to this chart alone, this system shows heavy usage (higher than 85%) at several times during the reporting period. If this pattern was consistent throughout several and longer reporting periods, it could indicate the need for performance tuning or even an upgrade. Because of the limited reporting time, it would be difficult to accurately gauge the need for an upgrade. We recommend monitoring your system 24 hours a day for a week in order to gather a wider range of data upon which to make decisions.
    Template Properties
  • On the Template menu, click Properties.
  • On the General tab, you’ll see that the selected Chart Type is Graph and the Timeline Type is Continuous.
  • Click the Graph tab.
  • Y Axis Labeling is defined as percent.
  • Graph Type is Line.
  • Scale is Percent.
  • Stacking is not enabled, as indicated by the lack of a check mark in the check box.
  • A Marker has not been defined.

    • Figure 5.7 Template Properties, Graph tab
  • Since this is a graph, skip the Table tab and click on the Thresholds tab. Neither a maximum nor a minimum threshold has been defined, as indicated by the lack of any check marks in the check boxes. Information regarding the Table tab can be found in "Table Properties".
  • Click the Data tab (see Figure 5.8).

    • Figure 5.8 Computation window in the Data tab
  • Group has been defined as all processor.
  • Element is % processor time.
  • Individual has been selected from the "All" group type.
  • Sign is defined as positive.
  • The data set does not have its Warnings enabled (expressed by the lack of <cc> in the computation window).
  • Click Warnings in the Add Options window.

    • Figure 5.9 Set Entry Warning Information dialog (example)
  • The Critical and Warning Level Thresholds are set to 0.000.
  • Click OK to close the Set Entry Warning Information dialog box.
  • Click the Links tab. The CPU Utilization chart is linked to the Processor Utilization by Category (NT) and Processor Queue Length (NT) charts. This means that by using the shortcut menu (right mouse click), you can access these charts.
  • Click Close at the bottom right of the Template Properties dialog box.
    • Modify Chart
  • On the Chart menu, click Modify Chart. Make sure that CPU Utilization is selected in the Chart drop-down menu.

    • Figure 5.10 Chart menu with Modify Chart selected
  • Click through the tabs. None of the local settings have been enabled.
  • Click OK to close the Modify Chart window.
    • Global Chart Options
  • On the Chart menu, click Global Chart Options.
  • On the Timeline Options tab:

    • Figure 5.11 Global Chart Options: Timeline Options tab
  • All is selected for the X Axis Range.
  • X Axis Points are set to Sample.
  • The Blanking Limit is 1441 minutes (24 hours).
  • No days are excluded.
  • Click the Global Date/Time Select tab.

    • Figure 5.12 Global Chart Options: Global Date/Time Select tab
    Note that:
  • Use File Start is enabled.
  • Use File End is enabled.
    • Because these are both enabled, we are currently seeing all the data that is available for this data file. From this point, we can only shorten the length of the reporting period displayed in our graphs, not expand it.
  • Click the Shift Settings tab (see Figure 5.13).

    • Figure 5.13 Global Chart Options: Shift Settings tab
    Note that:
  • Use All Week Setting is enabled.
  • Enable Shift Limits is selected. The shift has been defined as beginning at 00:00 and ending at 23:59.
  • Close the Global Chart Options dialog box.
    • Exception Window
  • On the View menu, click Exception Window.
    • The exception window shows that at 16 different points during the reporting period, the CPU utilization surpassed the Warning Level threshold ; and that at 8 points within that same time period, the CPU utilization surpassed the Critical Level threshold .
  • Close the Chart Exception window when you have finished (see Figure 5.14).

    • Figure 5.14 CPU Utilization Chart Exception window (example)
    Modify the CPU Utilization Chart
    To change how the graph looks and how it displays information, we are going to change the Template Properties, Modify Chart settings, and Global Chart Options. Remember that settings defined in the Global Chart Options affect all open charts but do not override any properties defined in the Modify Chart options. Modify Chart settings affect only the graph you select in the Chart pulldown menu on the General tab; and enabling any local settings in this menu overrides the template settings defined in the Template Properties dialog boxes.
    Template Properties
  • Click the Data tab. For this exercise, you’re going to change the Warning and Critical Level Thresholds. Highlight the data item:
    • (all processor (% processor time[Ind]))
  • Click Warnings in the Add Options window.
  • Change the Warning Level Threshold to 80.000.
  • Change the Critical Level Threshold to 90.000.
  • Click OK to close the Set Entry Warning Information dialog box.

    • Figure 5.15 Set Entry Warning Information dialog box
  • Click Apply to save your changes for the current session to the current template. Click Save As to save the altered template under a different name for use in future sessions.
  • Click Close to exit the Template Properties dialog box.
    • Modify Chart
    Since this graph is fairly dense, you can change the averaging to clear it up. Note, however, that by choosing to average the chart based on longer periods of time, it may become harder to pinpoint the exact moments or under what circumstances your system performance begins to suffer.
  • In the Chart menu, click Modify Chart.
  • Click the Averaging tab.
  • Click in the Local Averaging On check box.
  • Type 1 in the Minutes text box.
  • Click the Marker tab.
  • In the Marker section, click in the check box to enable a local horizontal marker.
  • Type 85 into the Marker text box. At this level (85%), you are using the CPU to it’s full potential. If the CPU usage is higher than 85%, it can lead to bottlenecks and slow data processing.
  • Click OK. You should see a considerable difference in the graph (see Figure 5.16). All the data points that had been collected at 15-30 second intervals have now been averaged into one minute interval data points. Also notice that now only three data points are above the 85% threshold line.

    • Figure 5.16 CPU Utilization graph with Averaging set to 1 minute and a marker at 85%.
    Global Chart Options
    Because there isn’t much change in the data points between 14:34 and 14:45, you can elect not to enable Use File Start, and instead begin the chart data at a point where the data begins shifting.
  • On the Chart menu, click Global Chart Options.
  • Click the Global Date/Time Settings tab.
  • Click in the Use File Start check box to disable it. The absence of a check mark indicates that function has been disabled. The Start Date and Start Time labels also become active (no longer grayed out).
  • Using the Start Time spin box Up Arrow, change the start time to 14:46 (see Figure 5.17).

    • Figure 5.17 Global Chart Options, Global Date/Time Select
  • Click Apply.
  • Click OK.
    • The graph is reformatted to show data points from 14:46 to 16:07.
    Exception Window
    On the View menu, click Exception Window.
    Because the warning and critical thresholds were added, this window now shows that at nearly 30 different points during the reporting period, the CPU utilization surpassed the Warning or Critical threshold limit. Close the Chart Exception window. Minimize the CPU Utilization chart when you have finished.

    Logical Disk Response Time Chart

    This chart displays how long it takes from the time a request for data (read or write) is made to the time the data is read from or written to the disk. Most requests can be satisfied within 11 ms (0.011 sec).
    Identify Chart Properties
    Because we made a global change to the CPU Utilization chart, all of the three remaining charts now show data from 14:46 to 16:07. To return the remaining charts to their original reporting periods:
  • On the Chart menu, click Global Chart Options.
  • Click the Global Date/Time Select tab.
  • Enable Use File Start by clicking in the checkbox. This causes all four of the graphs to return to the default start settings.
    • We also have a section of the Performance Gallery Gold window that is not being utilized. To fix this, do one of the following:
  • On the Window menu, click Tile Horizontal.
  • On the Window menu, click Tile Vertical.
  • On the Window menu, click Cascade.
    • You will see a difference in how the three charts are tiled (the CPU Utilization chart remains minimized). Choose whichever option you prefer.
    Graph Attributes
    Take a closer look at the chart. Maximize the Logical Disk Response Time chart.
    Figure 5.18 Logical Disk Response Time graph
  • The left side of the graph, (the y axis) shows the average elapsed time, in seconds, of I/O requests during the sample interval for each physical disk (C, D, and F).
  • The x axis displays the length of the reporting period. In this case, the reporting period is from 14:34 to 16:07.
    • System Performance Conclusions
    This system shows very long response times (longer than 0.025 seconds) for all three disks at several times during the reporting period. If this pattern was consistent throughout several and longer reporting periods, it could indicate the need for performance tuning or even an upgrade. Because of the limited reporting time and the small sampling intervals, it would be difficult to accurately gauge the need for an upgrade.
    Template Properties
    The following table displays the attributes for this chart found in the Template Properties dialog box.

    General tab
    Chart Type
    Graph
    Timeline Type