The I/O Usage system event measures potential versus used I/O to find I/O bandwidth bottlenecks and determine their scope. The event monitors the LUNs that TASM determines are most vulnerable to throughput issues and reacts as their I/O usage nears capacity. For example:
- Monitor 10% of targeted LUNs.
- Trigger the event if 1% of those LUNs are at 80% of their potential bandwidth.
- Average the data over 10 minutes.
- Send an alert email to the DBA.
A DBA can define multiple I/O Usage events to monitor different bandwidth thresholds. The event is based on the TASM event interval.
Guidelines
- Teradata recommends that customers begin with the default values of monitoring 10% of targeted LUNs and triggering when 1% of monitored LUNs exceed the usage threshold. Customers can change these values later as needed.
- Due to the type of data involved, only the Averaging qualification method is available. For more information on qualification methods, see Event Qualification.
- To increase the precision of detecting I/O bandwidth issues, increase the percentage of LUNs monitored (up to 50 LUNs) and increase the percentage of targeted LUNs to trigger from, up to 100%.
- To reduce how often the event triggers, increase the percentage of LUNs that triggers the event.
Considerations
- TDWM automatically adjusts I/O values for systems with Workload Management Capacity on Demand (WM COD). For example, if a system has 75% WM COD (uses ¾ of the system) and bandwidth is exhausted, the I/O Usage event reports usage at 100%. The system is using 100% of the I/O available to it.
- TDWM does not adjust I/O values for systems with Platform Metering Capacity on Demand (PM COD). DBAs need to adjust bandwidth percentages to reflect the PM COD value. For example, if the PM COD value is 80%, the site needs to react to that 80% threshold as if it were 100%.