Intelligent Dynamic Baseline Thresholds using Historical Data

I would like the ability to have a baseline that is intelligent and dynamic and adjustable so it takes into account day of the week/weekend, time of day, historical trend and sets it's threshold based on this information. For example, I have some items I'm monitoring that will increase and decrease based on the date and time and average utilization. On a Sunday at 3:00 AM there is next to no traffic so the baseline that is statically set for a median value will not be met and it would trigger an alarm. Conversely on a Saturday at 10:00PM the traffic could be through the roof so the baseline for the rest of the data samples might be too low for the current trend. I'd like to be able to adjust the number of days to take into consideration as well as +/- a couple of hours of sample data to set an intelligent dynamic threshold.

So in the example Saturday at 10:00PM I'd like to take into account the last X Saturdays Min, Max, Average, from 9:00PM - 11:00PM so that it has more data samples to take into consideration to build the trending data. Then it sets the threshold based on this information so it goes up and down with the traffic flow variations.

I would also like to be able to project expected threshold based on trends for example traffic has been increasing over the last few days or weeks seeing about at 10-15% increase in traffic. This would allow us to project for expansion or reductions in hardware/systems.

I've opened a case on this and was requested to post on THWACK to get support from the community on this feature request. Case # 00089156

Find more posts tagged with

threshold

historical

baseline

forecasting

intelligent

dynamic

Status: None

Comments

kpmarcin

Great idea! I had at least few customers over the year, which were asking this from me.

To SolarWinds Engineers, please consider implementing this.

xtraspecialj

Our company is moving more and more things to Dynatrace because of this. While Dynatrace is mainly a DevOps monitoring platform, its Infrastructure Monitoring does most of the important things that SolarWinds does too (CPU, Memory, Interfaces, VMWare, etc...), but it does it better. Its Dynamic Baseline Thresholding is super intelligent and takes a lot of things into consideration when creating a threshold. It compares day, time, and change percent into consideration before reporting anomalous behavior.

If SolarWinds doesn't implement more intelligent thresholding and anomalous detection soon, I can see our 4,000 servers completely leaving our SolarWinds environment and moving to Dynatrace using Infrastrucure-Only mode. Soon our SolarWinds environment will just have a few hundred network devices in it and that's it. We'd be dramatically dropping our licenses because of this too.

Heck, Dynatrace says they are adding in network monitoring soon as well, so we may just end up completely leaving SolarWinds. Yeah, Dynatrace is expensive, but dang is it good.