About data model acceleration

Splunk Analytics for Hadoop reaches End of Life on January 31, 2025.

For more information about how Data Model Acceleration works in Splunk Enterprise, see Accelerate data manuals in the Splunk Enterprise documentation.

Data model acceleration lets you create data models for data that includes virtual indexes. Splunk Analytics for Hadoop Data Model Acceleration uses cached information that can map extremely large datasets to accelerate searches.

Data model information that is stored in Splunk Enterprise indexes uses tsidx files. Splunk Analytics for Hadoop data models access data for virtual indexes that points to data in Hadoop, so you can create data models on any file types that a virtual index can point to. Splunk Analytics for Hadoop stores the data model acceleration files in Parquet and ORC.

For more information about how Data Model Acceleration works in Splunk Enterprise, see Accelerate data manuals in the Splunk Enterprise documentation.

How data model acceleration works in Splunk Analytics for Hadoop

1. You create a new data model Configure data model acceleration

2. Splunk Analytics for Hadoop creates a data model acceleration summary file for each raw data file:

• Splunk Analytics for Hadoop maintains information about the data model acceleration summary files in the KV Store (this allows fast lookups).

• Splunk Analytics for Hadoop stores the actual data model acceleration summary files in Hadoop.

3. In the case of a data model that covers a large time range, such as "one year" or "all time". You can partition data model files into buckets the same way you would in Splunk Analytics for Hadoop Report Acceleration. This makes both updating lookup files at model creation time and loading lookup files at search time faster. For more information see Accelerate Reports.

4. You can run a search on data in Splunk Analytics for Hadoop virtual indexes that can also include Splunk Enterprise indexes. If a data model exists for any Splunk Enterprise data, data model acceleration will be applied as described In Accelerate data models in the Splunk Knowledge Manager Manual.

5. For tstats/pivot searches on data models that are based off of Virtual Indexes, Splunk Analytics for Hadoop uses the KV Store to verify if an acceleration summary file exists for a raw data split.

6. If an acceleration summary file is not found for the raw data split, Splunk Analytics for Hadoop falls back to the original data file and does a regular search of the virtual index.

7. If an acceleration summary file is found, then the summary file is used instead of the raw data file.

8. The existing saved summary data is used to return data faster for a search that is comprised of all or part of that data model.