When does the accelerated Data model's values get processed to tsidx?

[clyde772]

Hey Experts!

We have a debate going on about when splunk punches in the field values when the data model acceleration is enabled. 2 ideas,

1. The accelerated field values gets processed (put into tsidx) when the data gets indexed into an index.
2. The accelerated field values gets processed (put into tsidx) when a user accesses a data model. Once the data model in accessed the processed values remains for the other users.

Which one is right? Also lookup, geo tag, regex attributes also gets put into accelerated tsidx?

Cheer~! Thanks for sharing your knowledge with people in need~!

[martin_mueller · Accepted solution]

If you're talking about persistent acceleration, number 1 is more correct - every few minutes splunk will accelerate recently indexed data on the indexers.
If you're talking about ad-hoc acceleration when you did not explicitly enable acceleration for this datamodel, number 2 is more correct - when you run a pivot, temporary acceleration happens on the search head for your second run in that pivot editor session. However, other users don't benefit from this because it's only temporary to your dispatch directory.

More info on that here: http://docs.splunk.com/Documentation/Splunk/latest/Knowledge/Acceleratedatamodels

Concerning your other types of fields, I'm not certain. For lookups and geoinformation I'm guessing no, because those use external sources that may change without splunk knowing. For regex I'm guessing yes, because those don't change on their own. However, that's just my gut talking.

[tchen_splunk]

I believe the geo-ip data is baked into the data model acceleration.

http://docs.splunk.com/Documentation/Splunk/6.2.0/Knowledge/AddageoIPattribute
Note: Geo IP attributes are added to your object as required attributes, and their Type values are predetermined. You cannot change these values.

[clyde772]

Martin, Appreciate your answer! Every few minutes? I verified I was able to get values for tstat up to the very second of current time. That means it would rather immdiately as the data gets indexed, wouldn't you think? This data model could be used very well to cover up Splunk regular data call speeds.