Skip to end of metadata
Go to start of metadata

You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 3 Next »

The Startup Time (ANALYZE > Graphs > Trend View > Startup Time) and Login Time (ANALYZE > Graphs > Trend View > Login Time) trend view graphs gives you overviews of periodical trends in your startup or login time data.
This graph is useful because ... it lets you detect if startup and login times are increasing in all or individual parts of your organization.
Have you ever wondered exactly when a computer is considered to be started, or exactly when a user is considered to be logged in? Read Startup and Login Measurement Facts to find out.
Data is aggregated on a per-period basis. The available periods depend on the PerformanceGuard configuration.

  1. Agents: Select the group of computers that the chart should be based on. Super groups, that is groups of groups, are enclosed in <>.
  2. Trend Period:

Example: If you select Day, the graph will display one data point per day.
An hour-based trend period is particularly useful when you view data from up to one week. A day-based trend period is useful when you view data from up to one month. A week-based trend period is useful when you view data from more than one month. A month-based trend period is useful when you view data from more than six months.

  1. Type: Select which type of information you want to see on the chart. You can choose averages or 90-percentiles for each trend period.

What's a 90-percentile?  It is the value below which 90% of the startup/login times are found. For example, if your startup times have a 90-percentile of 25 seconds, 90% of your computers start up in 25 seconds or less.

  1. Interval: Select the period of time that the graph should cover.

What does the interval X days mean?  If you select X days, you can define your own period, for example 90 days (which would cover the last 90 days). This can be particularly useful if you are going to include the trend graph in a report: Many organizations generate monthly trend reports that each cover the last 90 days. This would not always be possible with the Last Quarter interval, but with the X days interval you can easily generate such reports.
If the predefined intervals don't suit you, select Custom to specify your own interval.
Data is aggregated on a per-period basis. The available periods depends on the PerformanceGuard configuration (ADMINISTRATION > Reporting > Period).
Average values on trend graphs are based on a weighted average. That means that each value to be averaged is assigned a weight based on the number of occurrences of that value.
Computer Startup History
The Startup History tab shows login events captured by the PerformanceGuard agent installed on the computer.
To access the startup history, select ANALYZE > Computers > Computer Search, and search for a computer. In the search results, click the name of the required computer, and then select the Startup History tab.
Each startup is shown as a horizontal bar. The time (in seconds) it took the startup to complete is shown on the vertical axis of the graph.
You are able to adjust:

  1. The maximum number of startups shown on the graph
  2. The time interval that the graph covers

Below the graph you can view a list of the startups that have taken the longest time to complete.
Want detailed information about what goes on when the operating system is started on a computer? See Computer Startup Info.
Computer Startup Info
The Startup Info tab shows detailed information about what goes on when the operating system is started on a computer
To access the startup info, select ANALYZE > Computers > Computer Search, and search for a computer. Then, in the search results, click the name of the required computer, and then select the Startup Info tab.
Details for the last recorded startup of the computer is shown.
The first line in the table (OS Boot) isn't really a process, but it marks the first "life" of the operating system, just before the operating system starts to load drivers. The time of start for this event is set to zero, and the rest of the lines in the table have a time of start relative to this.
The table shows the following for each process:

  1. Id: Process ID.
  2. Time Start (s): Time when the process was started, measured in seconds relative to the start of the boot.
  3. Time Delta (s): The number of seconds that the process was running before the next process started. Only processes started by system accounts are considered, which means that user processes will always have a zero time delta.
  4. Process Name
  5. Clock Time (s): The total number of seconds that the process had been running at the time when the startup report was gathered.
  6. CPU Time: The amount of time that the process has used the computer's CPU, from the time it was started until the startup report was gathered.
  7. Disk IO (Reads): Number of disk reads performed by the process.
  8. Disk IO (Writes): Number of disk writes performed by the process.
  9. Other IO (Ops.): Number of non-disk IO operations performed by the process.
  10. Other IO (Bytes): Amount of bytes transferred by the process in non-disk IO.
  11. Page Faults: Total number of page faults that the process has generated.
  12. User: User and domain that owns the process.
  13. Version: Version of the process.
  14. Flags:
    1. 0: No information
    2. 1: The process is signed
    3. 2: The process signature is trusted
    4. 4: The process runs under Windows File Protection (WFP)
    5. 8: The process module has version information.
    6. 16: The process can be queried for process information
    7. 32: The process is a system process
    8. 64: The process hosts one or more services
    9. 128: The process was running when the report was created
  15. Hosted Services: Names of services that the process hosts (if any), see the following.

The values for CPU, IO and Page Faults are the accumulated number of events or bytes during the total run time of the process, that is during the time indicated in the Clock Time column.
Processes that start together with the operating system, and terminate before the PerformanceGuard agent service has been started won't appear in the list because no trace of them is left in the operating system.
Hosted Services
For each process that hosts one or more Windows services, the list of hosted services is shown in the last column of the table. This can be quite useful, because some processes, such as svchost.exe, run a lot of different services. With the information in the last column you can thus determine exactly which services use resources when the operating system starts.
Computer Login History
The login history shows login events captured by the PerformanceGuard agent on the computer in question.
To view the login history, select ANALYZE > Computers > Computer Search, and search for a computer. Then, in the search results, click the name of the required computer, and then select the Login History tab.
Each login is shown as a horizontal bar. The time (in seconds) it took the login to complete is shown on the vertical axis of the graph.
You are able to adjust:

  1. The maximum number of logins shown on the graph,
  2. The time interval that the graph covers

Below the graph you'll see a list of the logins that have taken the longest time to complete.
Computer Login Info
Login info lists login processes captured by the PerformanceGuard agent on an individual computer.
To view the login info, select ANALYZE > Computers > Computer Search, and search for a computer. Then, in the search results, click the name of the required computer, and then select the Login Info tab.
The table by default lists processes from the latest login. To view processes from another login, select the required login date and time in the menu in the top right corner of the table.
If a login took place some time ago, data about the login may no longer be available. This will depend on how long you keep process data for the agent configuration group in question (see information under Advanced in Agent Configuration Group Settings).
The table shows the following for each process:

  1. Id: Process ID.
  2. Time Start (s): Time when the process was started, measured in seconds relative to the start of the login.
  3. Time Delta (s): The number of seconds that the process was running before the next process started. Only processes started by system accounts are considered, which means that user processes will always have a zero time delta.
  4. Process Name
  5. Clock Time (s): The total number of seconds that the process had been running at the time when the startup report was gathered.
  6. CPU Time (s): CPU time, in seconds, that the process has spent.
  7. Disk IO (Reads): Number of disk reads.
  8. Disk IO (Writes): Number of disk writes.
  9. Other IO (Ops.): Number of non-disk IO operations.
  10. Other IO (Bytes): Amount of non-disk IO data.
  11. Page Faults: Total number of page faults that the process has generated.
  12. User: User and domain that owns the process.
  13. Version: Version of the process.
  14. Flags:
    1. 0: No information
    2. 1: The process is signed
    3. 2: The process signature is trusted
    4. 4: The process runs under Windows File Protection (WFP)
    5. 8: The process module has version information
    6. 16: The process can be queried for process information
    7. 32: The process is a system process
    8. 64: The process hosts one or more services
    9. 128: The process was running when the report was created
  15. Hosted Services: Names of services that the process hosts (if any). This can be quite useful, as some processes are used to run a lot of different services. This allows you to determine which services use resources during login.

Startup and Login Measurement Facts
PerformanceGuard can measure computer startup and login times ... but when do the measurements begin and end, and what are the success criteria? Find the answers here.
When Does Startup Measurement Begin?
PerformanceGuard begins to measure startup performance the moment that the computer's operating system begins its work. Technically, that's when the computer's OS Boot begins. OS Boot marks the first sign of life from the computer's operating system, just before the operating system begins to load drivers.
PerformanceGuard considers that point in time as zero, and it measures the beginning of subsequent startup processes relative to that point. This means that if you view Computer Startup Info in the PerformanceGuard web interface, the Time Start (s) column will indicate how many seconds after OS Boot that a given process began.
Thus, PerformanceGuard startup measurement doesn't begin the moment that a user switches on a computer. It begins a few moments after, when the computer's operating system begins its work.
OS Boot happens before the PerformanceGuard agent starts. How can you measure data before the agent starts?  Because the PerformanceGuard agent collects process information from the computer, it can easily find out which processes begin on the computer before the agent itself is started. The only exception is processes that start together with the operating system and terminate before the PerformanceGuard agent service has been started. Such processes won't be measured by the agent, because no traces of them are left in the operating system.
When Is a Computer Started?
With its default settings, PerformanceGuard considers a computer to be started when all of the following criteria are met:

  1. The process winlogon has started. winlogon is the operating system component that's used to authenticate a user who wants to log in to the computer. In other words, this criterion is met when the login dialog appears on the computer.
  2. The PerformanceGuard agent has started in local system context.

If required, PerformanceGuard administrators can change the startup success criteria for an agent configuration group: Select the required agent configuration group, and edit the agent configuration group setting called Startup Condition.
What's Measured During Startup?
During the startup of a computer, PerformanceGuard measures OS Boot (see When Does Startup Measurement Begin? in the previous) and every process that takes place until the startup success criteria are met (see When Is a Computer Started? in the previous).
The only exception is processes that start together with the operating system and terminate before the PerformanceGuard agent service has been started. Such processes won't be measured by the agent, because no traces of them are left in the operating system.
The following example shows startup process data from a real computer collected by a real PerformanceGuard agent and inserted into a chart tool. OS Boot marks the beginning of the startup measurement (zero), three processes start 15.968 seconds after OS Boot, yet another process starts 20.203 seconds after OS Boot, and so forth:

Note that the example simply shows the start time of each process relative to OS Boot; it doesn't contain end times for processes. In reality some processes may begin and end before others, like in the following example that almost resembles a Gantt chart that many people are familiar with from project plans:

Also note that when processes run simultaneously, they may affect each other.
Finally, note that some processes can host multiple Windows services. For example, the process svchost.exe runs several different services. If you looked closely at the first example, you probably noted that scvhost (short for service host) appeared several times, for example here:

In PerformanceGuard there are several ways in which you can find out which processes cause the longest delays during startup. For example, the Computer Startup Info in PerformanceGuard provides a good overview of process time measurements during startup of individual computers, including information about any services hosted by the processes.
When Does Login Measurement Begin?
With its default settings, PerformanceGuard begins to measure login performance the moment that the user's login credentials are validated and the login session begins.
If required, PerformanceGuard administrators can change the login measurement start condition for an agent configuration group: Select the required agent configuration group, and edit the agent configuration group setting called Login Start Condition.
When Is a User Logged In?
With its default settings, PerformanceGuard considers a login to be completed when the all of following criteria are met:

  1. The process explorer.exe has started. explorer.exe is the component that enables the user's desktop, the taskbar, file browsing, etc.
  2. explorer.exe uses less than 10% of the computer's CPU. In other words: explorer.exe has reached a stable state.
  3. The PerformanceGuard agent has started in user context.

If required, PerformanceGuard administrators can change the login success criteria for an agent configuration group: Select the required agent configuration group, and edit the agent configuration group setting called Login End Condition.
What's Measured During Login?
During the login, PerformanceGuard measures every process that takes place until the login success criteria are met (see When Is a User Logged In? in the previous).
In PerformanceGuard there are several ways in which you can find out which processes cause the longest delays during login. For example, the Computer Login Info in PerformanceGuard provides a good overview of process time measurements during login on individual computers.
Note that when processes run simultaneously, they may affect each other. Also note that some processes can host multiple Windows services.
Can a startup process continue during login?  Yes, many of the processes measured during startup will continue to run during login. Indeed, some of those processes may continue to run as long as the computer itself runs. The already running processes may of course affect the speed with which the computer is able to execute further processes, but that's standard behavior on a computer, and it can't be said to affect the validity of the PerformanceGuard login measurements.

Search this documentation

On this page

In this section

  • No labels