AWS prescriptive-guidance documentation change
Summary
Updated section headings and content ordering; expanded bullet points for clarity in the introduction; modified content about partition exchange and index removal; fixed typo in logging section heading
Security assessment
Changes focus on performance optimization techniques for write workloads (index management, data loading methods, MVCC cleanup). No security vulnerabilities, exploits, or security features are mentioned. Modifications are purely about database performance best practices.
Diff
diff --git a/prescriptive-guidance/latest/hyperscale-aurora-mysql/tune-write-workloads.md b/prescriptive-guidance/latest/hyperscale-aurora-mysql/tune-write-workloads.md index bf021d210..8b12f71d9 100644 --- a//prescriptive-guidance/latest/hyperscale-aurora-mysql/tune-write-workloads.md +++ b//prescriptive-guidance/latest/hyperscale-aurora-mysql/tune-write-workloads.md @@ -7 +7 @@ -Move referential integrityAvoid heavy primary keysUse partition exchangeRemove unused indexesPurge old row versionsLogging +Move referential integrity into the application layerAvoid using heavy primary keysRemove unused indexesUse partition exchange for loading data into partitioned tablesEnsure that old row versions are efficiently purgedEnsure logging doesn't cause additional contention @@ -21 +21 @@ Keep your primary keys light. The InnoDB storage engine appends the primary key -## Use partition exchange for loading data into partitioned tables +## Remove unused indexes @@ -23 +23 @@ Keep your primary keys light. The InnoDB storage engine appends the primary key -If you’re writing large sets of data into partitioned tables, the combination of [LOAD DATA FROM S3](https://docs.aws.amazon.com/AmazonRDS/latest/AuroraUserGuide/AuroraMySQL.Integrating.LoadFromS3.html) and [partition exchange](https://dev.mysql.com/doc/refman/5.7/en/partitioning-management-exchange.html) can improve performance, because the main table is not being accessed for the inserts. Partition exchange involves a data definition language (DDL), and it places a metadata lock on your table. Ensure that this is done when there are minimal or no queries running on the table so that the partition exchange DDL can obtain the metadata lock without waits. The exchange itself takes just milliseconds to complete. +[InnoDB](https://dev.mysql.com/doc/refman/8.0/en/innodb-storage-engine.html) optimizes its query plans based on the growth of your data, and it's a good idea to check for unused indexes in your Database and remove them. Unused indexes consume IO when the application writes data into a table. Check the list of unused indexes, and verify that they are not indexes that are used in rare situations, such as quarterly reports. Also, note that some indexes are used to enforce uniqueness or ordering and must also be considered. @@ -25 +25 @@ If you’re writing large sets of data into partitioned tables, the combination -## Remove unused indexes +## Use partition exchange for loading data into partitioned tables @@ -27 +27 @@ If you’re writing large sets of data into partitioned tables, the combination -[InnoDB](https://dev.mysql.com/doc/refman/8.0/en/innodb-storage-engine.html) optimizes its query plans based on the growth of your data, and it’s a good idea to check for unused indexes in your Database and remove them. Unused indexes consume IO when the application writes data into a table. Check the list of unused indexes, and verify that they are not indexes that are used in rare situations, such as quarterly reports. Also, note that some indexes are used to enforce uniqueness or ordering and must also be considered. +If you're writing large sets of data into partitioned tables, the combination of [LOAD DATA FROM S3](https://docs.aws.amazon.com/AmazonRDS/latest/AuroraUserGuide/AuroraMySQL.Integrating.LoadFromS3.html) and [partition exchange](https://dev.mysql.com/doc/refman/5.7/en/partitioning-management-exchange.html) can improve performance, because the main table is not being accessed for the inserts. Partition exchange involves a data definition language (DDL), and it places a metadata lock on your table. Ensure that this is done when there are minimal or no queries running on the table so that the partition exchange DDL can obtain the metadata lock without waits. The exchange itself takes just milliseconds to complete. @@ -44 +44 @@ In the InnoDB implementation of multiversion concurrency control (MVCC), when a -## Ensure logging doesn’t cause additional contention +## Ensure logging doesn't cause additional contention