Performance comparison between mybaits Batch ExecutorType and for_each Xml

问题

I have a list of records to be inserted into DB using my baits. Previously, my code is something like:

for(Item item :  items){
    sqlSession.insert("insert", item);
}

Using this method works but I find there is dynamical incremental DiskIO at Mysql server, due to number of items.

As I have little access to MySql configuration, and hope to resolve this high disk io issue, I find some possible solutions:

1. using ExecutorType.BATCH for sqlSession
2. insert multiple values within a single insert statement, such as: insert into Item values <foreach collection="list" item="item" index="index" separator=","> (#{item.a}, #{item.b}...) </foreach>

Does anyone has idea which one is suitable to solve my high diskIO problem and what the main difference between these two solution? thanks.

回答1:

Although it is difficult to predict the effect on the disk I/O of your server, method 1 (using ExecutorType.BATCH) is the recommended way.
It basically inserts a certain number of items at a time.
As it internally uses JDBC's batch operation API which is designed to handle many rows, you could benefit from optimizations by the driver.

Notes:

• Finding a proper 'batch size' is important. Please see the example code in another answer.
• In case of MySQL, adding rewriteBatchedStatements=true to the connection URL usually improves the performance significantly.

Method 2 (multi-row insert) executes a single statement binding all items at once.
It could cause memory issues.

来源：https://stackoverflow.com/questions/56034380/performance-comparison-between-mybaits-batch-executortype-and-for-each-xml