Versions Compared

Old Version 1

changes.mady.by.user Sarabjeet Kaur

Saved on

compared with

New Version Current

changes.mady.by.user DocReviewer

Saved on

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

Applies to: (tick) Kyvos Enterprise  (tick) Kyvos Cloud (SaaS on AWS) (tick) Kyvos AWS Marketplace

...

Kyvos uses query history and processes over time to create an aggregation strategy. As users continue to query, Kyvos learns which aggregates need to be built processed and the system slowly and gradually recommends deepening aggregates for a semantic model based on query pattern and data pattern analysis. The objective is to improve query performance under the constraint of keeping aggregates’ size (i.e. the total semantic model size) in control.

...

The Recommendations review includes suggesting an aggregation strategy. When you use Get Recommendations on a semantic model, you can recommend building processing only the aggregation strategy or only newer aggregates. When you get recommendations for an already processed semantic model, the system recommends newer aggregates and removes previous aggregates that are not in use.

...

Panel
panelIconIdatlassian-note
panelIcon:note:
bgColor#DEEBFF

Note

  • With the Kyvos 2023.1 release onwards, during the Update Aggregate process, if some cuboids are successful and others fail, the process is marked as a  Partial Success  with the percentage of successful cuboids displayed on the Job Summary. Additionally, the newly created cuboid is now marked as browsable. 

  • You can also view the log details of the input cuboid versus the output cuboid along with Measure Datasets by using the Download Logs option on the Job Summary page. 

  • On the Update Aggregate process summary screen, you will see various details, including updated partition details, advanced properties, error/warnings, and cuboid details when the semantic model is partially builtprocessed.

Aggregation Strategy

You can define an aggregation strategy when designing a semantic model. You can also specify recommendations for physical view, partitions, and advanced properties.

...

  • Expert (user-driven) is configuration-driven and allows you to tune your semantic model performance by setting up key materialization values and the precompute threshold level. The application displays recommendations based on historical queries and data profiles if needed. You can view how many dimensions or hierarchy materializations are selected from the summary screen. You can view details, make changes, or learn more by clicking thei symbol. You can also see the degree of materialization and the precompute threshold. For all of these items, you can see whether it is the default, inherited, or modified. Use this option if you want to have more control over what aggregates to be builtprocessed. 

  • Smart (system-driven) is query-based, allowing you to filter or analyze historical queries and work with data profile information. You can set the materialization level and specify whether to use fresh aggregates or to add additional aggregates in addition to any existing ones. Use this option if you want to get quick insights with a low learning curve. You will be prompted to regenerate aggregates after changes to the semantic model design.

Panel
panelIconIdatlassian-note
panelIcon:note:
bgColor#DEEBFF

Note

  • Switching from the smart aggregation strategy mode will clear existing aggregates on saving the semantic model. 

  • Changing the Aggregation Strategy Mode requires a full process. It may also result in the failure of subsequent incremental processes.

For processed semantic models, you can also use the Recommend me link to learn more. You will see the recommendations, descriptions, and reasons. At a glance, you can also see what will be added, modified, moved, or deleted. You can modify the recommendations by clicking the Actions menu (...) or click Accept. 

...