The calculations for metrics such as , , and include the use of the Network Latency metric. This metric is valuable in these calculations because it helps estimate the network delays that cannot be directly measured by the appliance from its server-side position on the network.
Network events are highly variable. Network events can arrive in tightly packed groups (also called bursts) followed by long delays before another burst of events occur. The size of these bursts are significant because they put the most load on the monitored network’s infrastructure. The appliance measures these events to determine the largest burst size that can occur using a group of metrics referred to as Peak Counts.
Service Start Peak Count tracks peak count of service start events.
Service Peak Count tracks peak count of service completion events.
It is possible that a burst of events could span the one-second interval. This could possibly result in a peak count that is less than the real maximum burst size. This issue is mitigated, however by the large number of samples (300) that are taken during a 5-minute interval. The large number of samples should ensure that an accurate measurement of the largest 1-second burst size is captured.
In a typical page request for main.html on in-house web servers, the calculation of Page End-to-End Time starts when the server receives the request, includes time spent on all in-house servers, and ends when the server responds. For more information, see .
When an in-house web server hands off a request to a CDN server, Foglight Experience Monitor cannot track the time spent on the request by the CDN servers. Therefore, the reported Page End-to-End Time becomes the time the in-house web server spends on the initial request for main.html before passing it on to a CDN server. This results in a partial page response.
For instrumentation instructions, see “Instrumenting web pages” in the Foglight Experience Monitor Installation and Administration Guide.
Once installed and configured, the appliance continuously monitors web traffic and performs analyses in real-time. The metrics generated by the appliance are organized into time series. A time series defines the time-based granularity of the metric. For example, a metric may reflect every event in the 8:00 AM hour, every event that occurred last Thursday, or every event that occurred last week.
The length of time data is kept for each time series depends on the size of the time intervals in the time series. Larger intervals are kept longer than smaller intervals (for example, monthly data is kept longer than real-time).
Real-time, hourly, daily, weekly, monthly are considered time series because once an interval in the time series is complete, all metrics for all resources in the current interval are saved. After the data is saved, no new updates can be made to the metrics in that interval. A time interval is considered complete when the next event has a timestamp for the following time interval. The metrics for all resources in the current time interval are stored in the database.
This time series contains the smallest intervals of the different time series. Real-time intervals are five minutes long. For each hour, 12 new intervals are created. Intervals are labeled by their starting time. For example, the real-time interval that spans 08:00.00 to 08:04.59 is labeled as 08:00. The data for real-time intervals is kept for one week.
Time intervals in this time series are 60 minutes long. For each day, 24 new intervals are created. Intervals are labeled by their starting time. For example, the hourly time interval that spans 13:00 to 13:59.59 is labeled 13:00. The data for hourly intervals is kept for one month.
Time intervals in this time series are 24 hours long. For each week, seven new intervals are created. Intervals are labeled by their common day of the week abbreviation (Sun, Mon, Tue, Wed, Thu, Fri, Sat). The interval boundary for a day is set at 00:00.00 till 23:59.59. The data for daily intervals is kept for one month.
Time intervals in this time series are seven days long (each day is 24 hours long). During a year, 52 new intervals are created. The interval boundary for a week goes from Sunday at 00:00.00 till Saturday at 23:59.59. Intervals are labeled by the starting date for each week (for example, 12/22). The data for weekly intervals is kept for two months.
Time intervals in this time series vary in length from 28 days to 31 days long (each day is 24 hours long). During a year, 12 new intervals are created. The interval boundary for a month goes from 00:00.00 on the first day of the month to 11:59.59 on the last day of the month. Intervals are labeled by the common abbreviation for each month (for example, Jan, Feb, Mar). The data for monthly intervals is kept for 12 months.
© 2020 Quest Software Inc. ALL RIGHTS RESERVED. Feedback Términos de uso Privacidad