Kaseya Community

trending usage

  • Can someone give me a practical example of how they use trending and what the best way to pull up the trending graphs and analysis is? Maybe use Disk space c: as the example?


    Legacy Forum Name: trending usage,
    Legacy Posted By Username: saybrook
  • There is a thread here if you want to continue the discussion. I, myself, have never used trending but here is a quote from the thread listed below.


    Matthew Bartels
    Well, what you've described is just how a monitor set works, nothing about trending specifically.

    So here's what it does, and my thoughts on when/why to use it:

    1. It does "linear regression" analysis on the last 2500 data points collected.

    2. It takes what you specify as the "trending window" to extend that linear approximation out however long... window of 1 day means it extends the line out 1 day into the future.

    3. It gives a "trending alarm" if the value *will* exceed the alarm threshold within the trending window, based on the linear approximation.

    Soooooo... when do you want to use this? Here are some examples of things you do *not* want to use trending on:

    * CPU utilization... it's up and down all day, never steady enough to get a linear approximation on anyway... just a simple threshold is fine (like "over 85% for 1 hour" on servers)

    * Uptime... well, it'd be a reliable indicator anyway Smile But there's no value - yes, it increases linearly, but it is so linear that if you want to alarm on uptime, just alarm on uptime!

    Now, places you might well want to use it, and examples of "sane" values:

    * The big one is "Disk Free Space" - alarm on "less than 10GB", and do trending over 6 hours... that way, if something starts steadily burning through your disk, you'll get a 6-hour headstart on fixing it. If whatever runaway process starts increasing the speed at which it's using disk space, well then all bets are off... nothing is perfect!

    * Memory utilization - you might consider this as well, although not on a SQL server, since the SQL process is *supposed to* grow to the size of RAM... maybe on a file/print, or DC it would make sense... but you'd probably want a much shorter window, like maybe 30 minutes or an hour, since memory utilization is much more transient.

    * Error rate - this would be possible, if you're watching email delivery errors, interface errors, etc... basically letting you have a little lead-time if the error rate starts shooting up, you might have a little better chance of fixing it before it starts impacting productivity. But often, these things are "go/no-go" so once you start throwing errors, you're already down... but it is a possibility.

    Hopefully this sheds some light on the topic!

    Legacy Forum Name: Monitor Sets,
    Legacy Posted By Username: GDRBrian