Advanced governance reporting provides functionality to aggregate algorithm execution metrics and make them searchable in order to easily view and report on resource usage at the algorithm, account, organization, and cluster level.
This feature is only available to users on Algorithmia Enterprise clusters where the feature has been enabled. The feature is available beginning in Algorithmia version 20.5.57.
- View historical information about system resource usage at the algorithm, account, and organization level.
- Support chargeback/showback efforts with report granularity down to the level of specific algorithm versions.
Governance reports contain both historical and aggregate data. Historical data is a roll-up from a time window of algorithm executions, while aggregate data represents a snapshot from the time the report is generated. The following metrics are available in governance reports:
- Algorithmia-hosted storage size (MiB)
- Data connector storage size (MiB)
- CPU core usage (duration of CPU core usage only, in seconds)
- CPU time (duration of total CPU-based execution in seconds, which includes CPU core usage)
- GPU time (duration of GPU-based execution in seconds)
- CPU cost
- GPU cost
- CPU memory usage (MiB)
- GPU memory (MiB)
- Algorithm execution count
- Algorithm error count
Aggregate data (aggregated at the organization level)
- Organization member count
- Organization-owned algorithm count
The governance reporting feature is available to cluster admins and organization admins. In order for cost information to be estimated, a cluster admin must first set expected resource cost rates. If the cost rates haven’t been set, all estimated costs will default to 0.
Cluster admins can view all resource consumption data captured within the platform, grouped by organization, account, and algorithm. They can also view everything that an organization admin can view, as described below. For more details, see the cluster admin docs.
Organization admins can view resource consumption data for organizations for which they’re an admin, for members of those organizations, and for algorithms owned by those organizations.
From the Reporting tab of an organization’s profile, data can be grouped by organization-owned algorithm or member account as needed.
Each metric listed above can be visualized by selecting it in the third dropdown, as shown above under Metrics. All metrics are also displayed in tabular form below the plot, ordered by the selected metric, in order to help you understand resource usage within a specified time period.
Organization admins can navigate to the Reporting tab of the account profile for an account that’s a member of the organization to see that specific account’s resource usage. At the account level, data can be grouped by organization or algorithm.
Organization admins can also navigate to the Reporting tab of the algorithm profile for an algorithm owned by their org to view information about usage for that specific algorithm. Algorithm usage metrics can be displayed at the organization level or grouped by the individual organization member account calling the algorithm.