On this page:
On this page:
Metric Generation and Time Zone Differences
In a distributed system, the metrics that the Controller collects may come from different time zones.
Before sending metrics to the Controller, agents normalize the times associated with the metrics by applying a UTC (Coordinated Universal Time) offset appropriate for their region.
In the Controller UI, all times are displayed in the Controller user's local time. To render local times, the UI applies the offset needed to render UTC time into the user's time zone as configured on the computer on which the browser is running.
While the time zone used in the UI is taken from the local environment and cannot be modified apart from changing the computer's system time, each UI user can modify the format of the time display, as described in User Preferences.
While the Controller UI displays times in the local system time, certain timestamps generated by AppDynamics reflect the Controller system time. These include, for example, event timestamps in notification messages, which reflect the Controller system time. Note that SaaS Controllers use Pacific Time (PT) as their system time.
Viewing Rolled Up Data in Graphs
Depending on your environment, AppDynamics measures hundreds, thousands, or even tens of thousands of metric values each minute. To display these results over time in a meaningful way, AppDynamics rolls up the data at regular intervals (for example, at 1-minute, 10-minute, and 1-hour intervals) based on how long ago the metric was measured.
In graphs throughout the AppDynamics user interface, you can hover over a point on the graph to get the value of the metric data measured during the interval represented by the point. The point represents the beginning of the interval. The length of the interval depends on the selected time range. Default settings are listed below.
- For up to 4 hours, data is rolled up every 1 minute, and displayed at 1-minute resolution. Each point on the graph represents a 1-minute interval; the point represents the beginning of the interval. Therefore, the data displayed for a 5:00 AM point represents the values measured from 5:00 AM to the millisecond before 5:01 AM.
- After 4 hours, data is rolled up every 10 minutes, and displayed at 10-minute resolution. Each point on the graph represents a 10-minute interval; the point represents the beginning of the 10-minute interval. Therefore, the data displayed for a 5:00 AM point represents the values measured from 5:00 AM to the millisecond before 5:10 AM.
- After 48 hours, data is rolled up every hour, and displayed at 1-hour resolution. Each point on the graph represents a 1-hour interval; the point represents the beginning of the hour. Therefore, the data displayed for a 5:00 AM point represents the values measured from 5:00 AM to the millisecond before 6:00 AM.
- After 365 days, data at 1-hour resolution is deleted.
Your administrator can adjust the amount of time data is retained at each resolution, although retaining data for longer than these default values is not generally recommended. See.
The following sections discuss the roll up periods and other time-related considerations when viewing metrics.
If you view data on a graph for any period in the previous 4 hours, each value you see represents a rollup of all the values measured during that minute. For example, if response time for an application is measured for 300 calls in a minute, the response time values of those 300 calls are averaged to present the Average Response Time (ART) for that minute. When you hover over a point on the graph, the Count figure represents how many times the metric data was measured during that minute.
In the graph below, response time was measured 140 times between 2:57 and 2:58 PM, and the ART for that minute was 889 ms.
In the following graph, the time range was set for the last three hours, so the interval is one minute. The errors per minute value displayed at the 1:34 PM point represents the number of errors recorded during the minute from 1:34 PM to the millisecond just before 1:35 PM.
From 4 to 48 hours, metric data is displayed at a 10-minute resolution. If you view data in a graph for a period that is between 4 and 48 hours old, each value you see represents a rollup of the values measured during that 10-minute period.
The ART for the rolled-up time period represents the true average. That is, the ART aggregates the sum and count of the total requests for the period; it does not average the ARTs calculated for the previous roll up granularity.
After 48 hours, metric data is displayed at a 1-hour resolution. If you view data in a graph for a period that is more than 48 hours old, each value you see represents a rollup of the values that were measured during that hour.
How the Last Interval is Displayed in a Graph
Because of the way rolled-up data is displayed, it may look as if data is not included for the last increment in the time range. However, the last data point in fact represents the last interval in the range.
For example, when you set a time range from 8 AM to 12 PM for a day that has been rolled up into 1-hour data points, you might expect to see five data points, one for 8 AM, 9 AM, 10 AM, 11 AM and 12 PM . However, the returned data consists of four data points for the hours 8 AM, 9 AM, 10 AM and 11 AM. The first data point, 8:00 AM, represents data collected from 8 AM to the millisecond just before 9 AM. The last data point, 11:00 AM, represents data collected from 11 AM to the millisecond just before 12 PM. So you are in fact seeing data from 8 AM to 12 PM, even though the last data point in the graph is 11 AM.
Viewing Graphs for Long Time Ranges
For time ranges that are greater than three weeks, graphs in the UI such as the Slow Transaction graph show the 14th of each month as the time point on the x-axis. For a time range that is less than two full months, this means that the graph may only have a single point, the middle of the month encompassed by the time range.
Viewing Details about Older Data
If you need to review details for an issue that occurred during a period for which you have only 10-minute or 1-hour data, AppDynamics provides access to diagnostic data via Transaction Snapshots. For example, suppose you are seeing values for 2 AM and 3 AM three days ago, but you need to examine details about a problem you were alerted about that occurred at 2:15 AM that morning. You view Transaction Snapshots for the hour between 2 AM and 3 AM to see details about what happened during that time period. By default, snapshots are retained for two weeks, and individual snapshots can be archived. See Transaction Snapshots.