Often questions come up on how to set up the trending options in the InetSupervisor. The settings reside in the Config table of the sql InetSupervisor database. You can access them by DabaseConfig then click on SqlDatabase then Config Table and finally click Load/Refresh.
Towards the bottom of the table you will see PropertyNames that begin with the word Trend. You can adjust their values to optimize the trend behavior. The starting point is property "Trend0Interval" – This is how often samples are taken from the network and stored as raw data in the InetTrends0 database. The Value is in minutes and the default is set to 15.
Next is the "Trends0ArchiveDuration" – This is how long the trend samples stored in the InetTrends0 database will be stored there before being compressed and moved to the next database InetTrends1.
Then you have same properties for Trends1, Trends2 and 3 and 4.
The key in selecting proper values is to keep the total number of trend samples in each database within these limits:
27 milion trend samples is the absolute limit where the SQL server express locks up and quits responding.
10 million trend samples is a limit you can run at on a fast RAID system with plenty of resources.
5 million trend samples is a more achievable limit with a moderate computer running current speed hard drive.
3 million trend samples or less is a limit we recommend on slower computers with 5400RPM hard drives.
How do we calculate total trend samples:
I-Interval in minutes , it's the TrendXInterval property from the Config table.
D- Duration in minutes, it's the TrendXArchiveDuration property from the Config table. This is a "Time To Live" since trend sample was taken.
C-Count, it's the total number of trended points
T-Total number of trend samples stored in a database.
So, we want to solve for T.
T = (1/I) * D * C
- Pick a target T total of points in the database – "number".
- Pick an interval I you want to gather trends at - "minutes".
- Look up the count C of the trended points. You can use Database Config Points table , then use a filter to select only trended points. At the top you will see the total number of items found.
- Calculate archive duration D =(T * I)C and plug it into the TrendArchiveDuration.
Repeat this process for each Trend database ie. Trend1, Trend2, … Trend4.
You want to increase the TrendXInterval each time you move to a higher database by a factor of 10 or more if possible. The last database should substantially increase the TrendInterval so you can achieve a long Archive Duration and therefore keep highly compressed data for a long time.
Example 1:
Low-end computer, interval trend frequency is I = 15 minutes, total number of trended points is C = 100. We would like a light load in the trend databases so we pick T = 1,000,000 trend samples.
D = (T * I) / C = (1,000,000 * 15) / 100 = 150,000 minutes = 104 days , so put in the following values into the config table:
- 15 for Trend0Interval
- 150,000 for Trend0ArchiveDuration
When values get older than 104 days I would like to compress them so the "resolution" interval I = 30 min.
D = (T * I) / C = (1,000,000 * 30) / 100 = 300,000 minutes = 208 days , so plug in the following:
- 30 for the Trend1Interval
- 450,000 for the Trend1ArviceDuration = (104 days in Trends0 plus 208 days in Trends1)
When values get older than 312 days then I would like to compress then so the "resolution" interval I = 60 min.
D = (T * I) / C = (1,000,000 * 60) / 100 = 600,000 min = 416 days, so put in the following:
- 60 for the Trend2Interval
- 1,050,000 for the Trend2ArchiveDuration = (104 days + 208 days + 416 = 729days)
When values get older than 729 days I would like to compress them so the Interval I = 180 minutes.
D = (T * I) / C = (1,000,000 * 180) / 100 = 1,800,000 min = 1250 days, so put in the following:
- 180 for the Trend3Interval
- 2,850,000 for the Trend3ArchiveDuration = (104+208+416+1250 days)
Finally we can stretch the very old data to I = 360 min = 6 hours.
D = (T * I) / C = (1,000,000 * 360) / 100 = 3,600,000 min = 2500 days, so put in the following:
- 360 for the Trend4Interval
- 6,450,000 for the Trend4ArchiveDuration = (104+208+416+1250+2500 days = 4478 days = 12 years)
End of Example 1.
Example 2:
We have C=1000 points to be monitored and would like to store samples every I=1 minute. We will load the databases with say T=2,000,000 samples each.
D = (T * I) / C = (2,000,000 * 1) / 1000 = 2000min = 1.38 day, therefore this goes into the Config table:
- 1 for the Trend0Interval
- 2000 for the Trend0ArchiveDuration (1.38 days)
When samples get older than 2000 min we compress them into 5 minute averages like so:
D = (T * I) / C = (2,000,000 * 5) / 1000 = 10,000 = 6.94 day, therefore this goes into the Config table:
- 5 for the Trend1Interval
- 12,000 for the Trend1ArchiveDuration (1.38 + 6.94 days)
When samples get older then 12,000min (8.33 days) we compress them into 15 min averages like so:
D = (T * I) / C = (2,000,000 * 15) / 1000 = 30,000 min = 20.83 days, therefore this goes into the Config table:
- 15 for the Trend2Interval
- 42,000 for the Trend2ArchiveDuration (1.38 + 6.94 +20.83 days)
When samples get older then 42,000min (8.33 days) we compress them into 60 min averages like so:
D = (T * I) / C = (2,000,000 * 60) / 1000 = 120,000 min = 83.3 days, therefore this goes into the Config table:
- 60 for the Trend3Interval
- 164,000 for the Trend3ArchiveDuration (1.38 + 6.94 +20.83 + 83.3 days = 113 days)
When samples get older than 164,000 min (113 days) we compress them into 60 min averages like so:
D = (T * I) / C = (2,000,000 * 360) / 1000 = 720,000 min = 500 days, therefore this goes into the Config table:
- 360 for the Trend4Interval
- 882,000 for the Trend4ArchiveDuration (1.38 + 6.94 +20.83 + 83.3 + 500days = 612.5 days = 1.7 years)
End of Example 2.
http://www.inetsupervisor.com/Support/default.htm "InetSupervisor SUPPORT"
http://www.inetsupervisor.com/Products/Default.htm "Quark Communications PRODUCTS"
http://www.inetsupervisor.com "InetSupervisor HOME"
Cheers;
Adam G.