How to stop data spikes or outliers appearing in CA Performance Management (CAPM) reports & Views

book

Article ID: 208054

calendar_today

Updated On:

Products

CA Performance Management - Usage and Administration CA Infrastructure Management

Issue/Introduction

Sometimes devices send back invalid data to polls or counters have rolled over before a reset. When this occurs, it can appear as a massive spike that diminishes or obscures all other data points. For example:

When this occurs, the data is invalid and the report rendered useless.

Environment

DX NetOps CAPM 20.2.4 and later

Resolution

There is no way to actually eliminate a single poll value.

However, there is a new feature to configure counters values that was introduced in 20.2.4 that can catch and drop spikes or outliers.

This is detailed in our TechDocs:

TechDocs : DX NetOps 20.2.x : CAPM - Counter Rollover Log

For example, adjusting the values for 

largeDeltaValueThreshold32
largeDeltaValueThreshold

When a device gives us a bad data spike, CAPM doesn’t normally drop it, its written to the Vertica DB along with all other poll data.

For 64 bit counters, we only drop spikes if they are >= 9,223,372,036,854,775,808.
For 32 bit counters, we only drop spikes if they are >= 4,294,967,296.

So you could configure the largeDeltaValueThreshold32 or largeDeltaValueThreshold depending on if that interface is low speed or high speed. You could use something like 100000000000 for largeDeltaValueThreshold.

This will remove such outliers/spikes from the graphs.

Attachments