Benchmarks for performance optimization

Updated 

Recommendation

By setting and observing benchmark maximums for key model elements, you can improve your user experience in areas such as navigation and readability, as well as avoid overtaxing your system resources.

Impact

Follow these recommendations to maximize the scalability and adoption of your implementation by improving performance, reducing the size of backup/restore files and time-consuming import and export queries. Equally, user interface and experience are improved by greater readability and navigation, and by supporting governance processes that are more practical and easier to sustain.

Best practices

Domain-level recommendations

  • Keep the number of domains within a single community below 1,000 to aid navigability for users.

    • Use ownership, stewardship, or governance councils as a basis for dividing communities with more domains into multiple communities.

  • Try to keep the total number of domains in your model below 10,000, as any more may make the model difficult to manage and navigate for users.

    • Use any business dimension, such as Line of Business, geographic region or data domain, as a logical basis to consolidate the number of domains. For example, consolidate all customer schemas or product schemas into a single domain.

Asset-level recommendations

  • The number of attributes per asset should not exceed 500. Beyond this limit, it becomes extremely difficult for users to read or navigate what distinguishes one asset from another. As with the benchmarks above, consider a logical or business dimension that will allow you to consolidate attributes.

  • Automatic hyperlinking of assets is turned off by default, but it can be turned on. However, if you allow the number of automatically hyperlinked assets to exceed one million, it can slow performance and be a negatively impact user experience and adoption.

  • If the number of responsibilities, direct or inherited, per asset exceeds 100, governance of the asset becomes difficult to sustain and navigate, and increases the risk of inadvertent conflicts among the asset’s responsibilities. A process with tens of roles involved often represents over-engineering of the governance process.

User recommendations

  • Exceeding 20,000 users per user group can lead to performance degradation, therefore use any business dimension, such as Line of Business, geographic region or data domain, as a logical basis to split large user groups into smaller groups.

Validation criteria

Review the elements above periodically to ensure you are not exceeding the suggested maximums. You can develop a custom workflow to capture volumes of the above and/or use Insights reporting.

Additional Information

Go to the diagnostics section in Collibra's Documentation Center for more information on any of the above elements.