Tuning Options
Tuning Options allow you to control how OtterTune interacts with your database for Knob Tuning Recommendations.
- Auto-Tuning: OtterTune will apply recommendations on your behalf
- Manual Review: You approve recommendations before allowing OtterTune to apply
- Self-Directed Tuning: You take our recommendations and apply them outside of OtterTune
AWS RDS exposes the database knobs that OtterTune tunes in the form of Parameter Groups. If you’d like OtterTune to tune your existing parameter group, we’d recommend first copying the parameter group to a new parameter group. This will allow you to switch back to your old, stable parameter group that OtterTune will not tune. Please see the AWS RDS Parameter Group documentation for more details.
The Parameter Group ARN can be constructed with the following format:
arn:aws:rds:<region>:<account>:pg:<parameter_group_name>(Aurora only) While Instance-level Parameter Groups have the same format as above, Cluster-level Parameter Groups have a different format:
arn:aws:rds:<region>:<account>:cluster-pg:<cluster_parameter_group_name>OtterTune cannot tune the default parameter group, you must select a custom parameter group.
The Aurora Tuning Level specifies whether OtterTune will tune the instance-level or cluster-level parameter group for your AWS Aurora DB cluster. Please see the AWS Aurora documentation for more details.
OtterTune 2.0 blends knob tuning recommendations that are machine learning-generated with recommendations curated from our exceptional team of database experts. Here you can specify whether you want to have your database recommendations generated only from heuristic means or with the blend of machine learning and human expertise.
The Target Objective is what OtterTune will optimize for when it generates new configurations for the database. You can change the Target Objective to guide OtterTune towards optimizing performance versus cost.
MySQL
PostgreSQL
Objective | Description |
|---|---|
Throughput | The number of transactions that the database completes. OtterTune computes MySQL's throughput by adding together the following metrics from MySQL's INNODB_METRICS internal table: trx_rw_commits+ trx_ro_commits+ trx_nl_ro_commits. |
AverageLatency | The average latency (in milliseconds) for disk operations as reported by Amazon CloudWatch. OtterTune computes this by adding together CloudWatch's ReadLatency + WriteLatency |
P99QueryLatency | The estimated P99 query latency calculated from query digest. We only collect query IDs and latencies. We do not collect query details, tables or schemas. |
Objective | Description |
|---|---|
Throughput | The number of transactions that the database completes. OtterTune computes PostgreSQL's throughput from the xact_commitmetric in thepg_stat_databasestatistics table. |
AverageLatency | The average latency (in milliseconds) for disk operations as reported by Amazon CloudWatch. OtterTune computes this by adding together CloudWatch's ReadLatency + WriteLatency |
P99QueryLatency | The estimated P99 query latency calculated from query digest. We only collect query IDs and latencies. We do not collect query details, tables or schemas. |
It is possible to configure OtterTune to optimize other objectives in your database, such as CPU utilization or storage IOPs. Please contact us to discuss if you need this capability.
When OtterTune’s knob tuning mode is set to generate machine learning-based recommendations, you’ll have the ability to control which knobs to tune and their allowed values.
OtterTune maintains a list of recommended knobs for each database type (e.g., PostgreSQL, MySQL) that it has learned will have the most positive impact on a database's performance. These knobs are listed in the "Recommended" section of the Tuning Knobs panel. All recommended knobs will be selected by default the first time you switch the Tuning Mode to "Optimizing".
Allowed values: OtterTune also suggests ranges for recommended knobs that are based on the database's original con instance's available hardware resources, the database's original knob settings, and domain knowledge.
The recommended knobs and allowed values are guidelines and may not be appropriate for all operating environments. We highly recommend that you review them thoroughly before tuning your database to avoid performance degradation or downtime due to misconfiguration.
All other knobs that OtterTune can tune are listed in the "Other knobs" section (visible after you scroll down past all recommended knobs). Note that there are some knobs that OtterTune cannot optimize. See our FAQ for details.
Allowed values: The ranges for other knobs are set to the minimum and maximum values permitted by the database system.
If you choose to tune other (non-recommended) knobs, it is very important that you select allowed values that are appropriate for your database and operating environment. For resource knobs, be sure that the maximum values you choose do not exceed the available resources for the instance.
Last modified 2mo ago