Sheet and Visualization Design Best Practices

Visualizations are organized into groups, which are organized into sections, which are on a sheet.

• Define the purpose of the sheet. When will an administrator use this sheet? How will an administrator use this sheet to learn more about this application or subsystem?

  A sheet should be self explanatory and guide the administrator through the troubleshooting process. A sheet should help an administrator discover contributors to a particular problem and lead the administrator to specific actions.

• Determine the information needed to achieve the purpose. What information is most key? What information is related?

  Resource utilization and saturation and number and type of errors typically are key information. Can you show utilization rather than raw usage numbers?

  Time-series visualisations provide the most data dense visualisation that include historical data. Other types of visualizations should only be used if the time-series visualisation does not suit the underlying data.

  Consider using //:op.top to avoid switching to Pareto visualization automatically when defining a time-series visualization.

  Are historical comparisons of data values useful? If so, should historical comparisons be shown by default?

  Is the required information available from existing statistics, do new statistics need to be created, do existing statistics need to be partitioned or mapped to achieve the purpose of this sheet?

• Determine the layout of the sheet into sections and groups. What flow should the administrator likely to follow to troubleshoot a specific problem or explore potential problems with this application or subsystem?

  • The most key information should be first, at the top of the sheet. To aid troubleshooting, provide information about potential problems or errors first, instead of starting with basic status information.

  • Sections should guide the flow of problem investigation. Based on what the most key information in the first section shows, what should the administrator look at next? Group the key indicators for the most common symptoms of problems with this subsystem into sections.

  • Related information should be in the same group.

  • Find the right balance. Organizing data into multiple groups and sections can make data easier to understand and use and reduce clutter on the screen, but too many groups and sections add to the clutter and make problems harder to investigate. Also, each section and group title takes screen real estate that visualizations could use.

• Use the purpose of the sheet to create a brief description of the sheet.

  • What data does this sheet provide?

  • What kinds of failures can this sheet help diagnose?

  • If possible, mention diagnosis specifics such as how visualizations are related and used together. What are typical steps to take to troubleshoot the system represented by this sheet?

  • Include links to additional information.

• Create a brief description for each section.

  • What is the purpose of this section? Which symptoms can be diagnosed here?

  • What steps should an administrator take here?

  • Include links to additional information.

• Create a brief description for each visualization.

  • How can the data in this visualization be used to diagnose a problem?

  • How can an administrator compare the data values in this visualization to expected values?

  • Mention specifics such as how data is partitioned, what related data can be shown, what you can learn from this data.

  • Include links to additional information to help troubleshoot the problem, including references to visualizations on other sheets and how those would be helpful.

  If you cannot clearly express the purpose of the visualization, perhaps the visualization is not needed to investigate problems with this subsystem.

• Decide what type of visualization will make the data easiest to understand and use.

  • Add operations to the SSIDs to show the most meaningful data. Do you want to show resource utilization, resource saturation, error counts, rate, top five values? Do you want to filter the data or convert to a different unit of measurement?

  • How important is historical data? To show historical data, use one of the time series type of visualizations. Consider whether the data identifiers should be in an enabled collection and collected persistently.

  • Is a stacked time series or Pareto chart better than a regular time series graph for this data? Is a gauge, bar chart, or pie chart better for what you want to show? Do you need a histogram to show which values occur most frequently?

  • Is data that is represented by different SSIDs related closely enough to show in the same visualization? For example, reads and writes of the same resources might be shown in a single visualization. To show two different SSIDs in one visualization, both SSIDs must be partitioned identically (or not partitioned), must be in the same units in a similar range of possible values, and must be able to be shown at the same time scale.

  • Are certain events useful to include with this data? Events can be shown as points on a time graph.