diff --git a/streamingpro-mlsql/src/main/java/streaming/core/datasource/impl/MLSQLDirectJDBC.scala b/streamingpro-mlsql/src/main/java/streaming/core/datasource/impl/MLSQLDirectJDBC.scala
index 2f71d382..9faf5d10 100644
--- a/streamingpro-mlsql/src/main/java/streaming/core/datasource/impl/MLSQLDirectJDBC.scala
+++ b/streamingpro-mlsql/src/main/java/streaming/core/datasource/impl/MLSQLDirectJDBC.scala
@@ -1,11 +1,11 @@
 package streaming.core.datasource.impl
 
 import java.util.Properties
-
 import com.alibaba.druid.sql.SQLUtils
 import com.alibaba.druid.sql.repository.SchemaRepository
 import com.alibaba.druid.sql.visitor.SchemaStatVisitor
-import com.alibaba.druid.util.JdbcConstants
+import com.alibaba.druid.util.{JdbcConstants, JdbcUtils}
+import org.apache.spark.{MLSQLSparkConst, SparkCoreVersion}
 import org.apache.spark.sql.catalyst.plans.logical.MLSQLDFParser
 import org.apache.spark.sql.execution.WowTableIdentifier
 import org.apache.spark.sql.mlsql.session.{MLSQLException, MLSQLSparkSession}
@@ -53,17 +53,36 @@ class MLSQLDirectJDBC extends MLSQLDirectSource with MLSQLDirectSink with MLSQLS
 
   override def load(reader: DataFrameReader, config: DataSourceConfig): DataFrame = {
     val format = config.config.getOrElse("implClass", fullFormat)
+    var driver: Option[String] = config.config.get("driver")
     var url = config.config.get("url")
     if (config.path.contains(dbSplitter)) {
-      val Array(_dbname, _dbtable) = config.path.split(toSplit, 2)
+      val Array(_dbname, _) = config.path.split(toSplit, 2)
       ConnectMeta.presentThenCall(DBMappingKey(format, _dbname), options => {
         reader.options(options)
         url = options.get("url")
+        driver = options.get("driver")
       })
     }
-
     //load configs should overwrite connect configs
     reader.options(config.config)
+    assert(url.isDefined, s"url could not be null!")
+    assert(driver.isDefined, s"driver could not be null!")
+    if (JdbcUtils.isMySqlDriver(driver.get)) {
+      /**
+       * Fetch Size It's a value for JDBC PreparedStatement.
+       * To avoid data overload in the jvm and cause OOM, we set the default value to Integer's MINVALUE
+       */
+      MLSQLSparkConst.majorVersion(SparkCoreVersion.exactVersion) match {
+        case 1 | 2 =>
+          reader.options(Map("fetchsize" -> "1000"))
+        case _ =>
+          reader.options(Map("fetchsize" -> s"${Integer.MIN_VALUE}"))
+      }
+
+      url = url.map(x => if (x.contains("useCursorFetch")) x else s"$x&useCursorFetch=true")
+        .map(x => if (x.contains("autoReconnect")) x else s"$x&autoReconnect=true")
+        .map(x => if (x.contains("failOverReadOnly")) x else s"$x&failOverReadOnly=false")
+    }
 
     val dbtable = "(" + config.config("directQuery") + ") temp"
 
diff --git a/streamingpro-mlsql/src/main/java/streaming/core/datasource/impl/MLSQLJDBC.scala b/streamingpro-mlsql/src/main/java/streaming/core/datasource/impl/MLSQLJDBC.scala
index 9b51f810..3ca95228 100644
--- a/streamingpro-mlsql/src/main/java/streaming/core/datasource/impl/MLSQLJDBC.scala
+++ b/streamingpro-mlsql/src/main/java/streaming/core/datasource/impl/MLSQLJDBC.scala
@@ -19,11 +19,12 @@
 package streaming.core.datasource.impl
 
 import java.util.Properties
-
 import _root_.streaming.core.datasource.{SourceTypeRegistry, _}
 import _root_.streaming.dsl.mmlib.algs.param.{BaseParams, WowParams}
 import _root_.streaming.dsl.{ConnectMeta, DBMappingKey, ScriptSQLExec}
 import _root_.streaming.log.WowLog
+import com.alibaba.druid.util.JdbcUtils
+import org.apache.spark.{MLSQLSparkConst, SparkCoreVersion}
 import org.apache.spark.ml.param.{BooleanParam, LongParam, Param}
 import org.apache.spark.sql._
 import org.apache.spark.sql.execution.datasources.jdbc.JDBCOptions
@@ -45,31 +46,50 @@ class MLSQLJDBC(override val uid: String) extends MLSQLSource with MLSQLSink wit
   def toSplit = "\\."
 
   override def load(reader: DataFrameReader, config: DataSourceConfig): DataFrame = {
-    var dbtable = config.path
+    var dbTable = config.path
     // if contains splitter, then we will try to find dbname in dbMapping.
     // otherwize we will do nothing since elasticsearch use something like index/type
     // it will do no harm.
     val format = config.config.getOrElse("implClass", fullFormat)
+    var driver: Option[String] = config.config.get("driver")
     var url = config.config.get("url")
     if (config.path.contains(dbSplitter)) {
-      val Array(_dbname, _dbtable) = config.path.split(toSplit, 2)
+      val Array(_dbname, _dbTable) = config.path.split(toSplit, 2)
       ConnectMeta.presentThenCall(DBMappingKey(format, _dbname), options => {
-        dbtable = _dbtable
+        dbTable = _dbTable
         reader.options(options)
         url = options.get("url")
+        driver = options.get("driver")
       })
     }
     //load configs should overwrite connect configs
     reader.options(config.config)
+    assert(url.isDefined, s"url could not be null!")
+    assert(driver.isDefined, s"driver could not be null!")
+    if (JdbcUtils.isMySqlDriver(driver.get)) {
+      /**
+       * Fetch Size It's a value for JDBC PreparedStatement.
+       * To avoid data overload in the jvm and cause OOM, we set the default value to Integer's MINVALUE
+       */
+      MLSQLSparkConst.majorVersion(SparkCoreVersion.exactVersion) match {
+        case 1 | 2 =>
+          reader.options(Map("fetchsize" -> "1000"))
+        case _ =>
+          reader.options(Map("fetchsize" -> s"${Integer.MIN_VALUE}"))
+      }
+
+      url = url.map(x => if (x.contains("useCursorFetch")) x else s"$x&useCursorFetch=true")
+        .map(x => if (x.contains("autoReconnect")) x else s"$x&autoReconnect=true")
+        .map(x => if (x.contains("failOverReadOnly")) x else s"$x&failOverReadOnly=false")
+    }
 
     val table = if (config.config.contains("prePtnArray")){
       val prePtn = config.config.get("prePtnArray").get
         .split(config.config.getOrElse("prePtnDelimiter" ,","))
 
-      reader.jdbc(url.get, dbtable, prePtn, new Properties())
+      reader.jdbc(url.get, dbTable, prePtn, new Properties())
     }else{
-      reader.option("dbtable", dbtable)
-
+      reader.option("dbtable", dbTable)
       reader.format(format).load()
     }