diff --git a/src/loader/DDBHFTLoader.py b/src/loader/DDBHFTLoader.py
index 2a02166..c14f786 100644
--- a/src/loader/DDBHFTLoader.py
+++ b/src/loader/DDBHFTLoader.py
@@ -119,12 +119,15 @@ class DDBHFTLoader(DDBLoader):
 
         # 不能重复创建Pool对象，因此需要在循环的最外侧创建好Pool对象，然后传参进去
         with Pool(self.num_workers if num_workers is None else num_workers) as pool:
-            for hft_type_name in self.hft_type_list:
-                print('Will work on hft type:', hft_type_name)
-                with tqdm(stock_list) as pbar:
-                    for stock_id in pbar:
-                        pbar.set_description(f"Working on stock {stock_id}")
-                        self.dump_hft_to_ddb(hft_type_name, stock_id, pbar=pbar, pool=pool)
+            # Always reuse the connection object, to reduce the memory consumption. 
+            with self.mssql_engine.connect() as conn:
+                # Loop through the stock list.
+                for hft_type_name in self.hft_type_list:
+                    print('Will work on hft type:', hft_type_name)
+                    with tqdm(stock_list) as pbar:
+                        for stock_id in pbar:
+                            pbar.set_description(f"Working on stock {stock_id}")
+                            self.dump_hft_to_ddb(hft_type_name, stock_id, conn, pbar=pbar, pool=pool)
 
 
     def _get_stock_date_list(self, cache=False):
@@ -354,7 +357,7 @@ class DDBHFTLoader(DDBLoader):
         print('-' * 80)
 
     
-    def dump_hft_to_ddb(self, type_name, stock_id, trade_date=None, pbar=None, pool=None):
+    def dump_hft_to_ddb(self, type_name, stock_id, conn, trade_date=None, pbar=None, pool=None):
         if (type_name, stock_id, 'OK') in self.dump_journal_df.index:
             message = f"Will skip ({type_name}, {stock_id}) as it appears in the dump journal."
             if pbar is None:
@@ -376,53 +379,52 @@ class DDBHFTLoader(DDBLoader):
         
         # 经过尝试，按个股来做batch查询效率还是可以接受的
         # mssql中，索引字段是(S_INFO_WINDCODE, TRADE_DT)
-        with self.mssql_engine.connect() as conn:
-            stat = """
-                select * from [Level2Bytes{mssql_type_name}].dbo.[{mssql_type_name}] 
-                where S_INFO_WINDCODE='{stock_id}'
-            """.format(
-                mssql_type_name = self.mssql_name_dict[type_name],
-                stock_id = stock_id
-            )
-            row_list = list(conn.execute(stat).fetchall())
-
-            # 如果`_journal_dt`不为空，则说明之前的日志中表明改股票数据已经部分完成，需要逐个核对日期
-            # 这里只把日期值不再`_journal_dt`的记录放入`row_list`
-            if _journal_dt is not None:
-                row_list = [row for row in row_list 
-                        if pd.to_datetime(row[1]) not in _journal_dt.index]
-                print(f"Resume job for {stock_id}, with {len(row_list)} rows left.")
-    
-            num_rows = len(row_list)
-            # 如果行数为0，则说明是空数据，可以直接返回
-            if num_rows == 0:
-                return
-
-            if pbar:
-                #pbar.set_description(f"Did get the result set for stock {stock_id} from mssql")
-                pbar.set_description(f"Will work in paralle on dumping job on {stock_id} of len {num_rows}")
-            else:
-                print(f"Did get the result set for stock {stock_id} from mssql")
-
-            # 每一行是当个个股某一日的所有高频交易信息
-            # 使用多进程来加快速度
-            
-            #with Pool(self.num_workers if num_workers is None else num_workers) as pool:
-            if pool is None:
-                print("Will create new Pool object, but this is not encourage for large batch work.")
-                pool = Pool(self.num_worker)
-
-            # 在单个股票内部，对不同日期进行并行处理，对内存使用较为友好，不需要同时载入多个股票海量的全历史数据
-            with tqdm(total=num_rows, leave=False) as sub_pbar:
-                for _ in pool.imap_unordered(
-                    functools.partial(
-                        DDBHFTLoader.dump_stock_daily_to_ddb, 
-                        type_name = type_name,
-                        stock_id = stock_id
-                    ),
-                    row_list
-                ):
-                    sub_pbar.update()
+        stat = """
+            select * from [Level2Bytes{mssql_type_name}].dbo.[{mssql_type_name}] 
+            where S_INFO_WINDCODE='{stock_id}'
+        """.format(
+            mssql_type_name = self.mssql_name_dict[type_name],
+            stock_id = stock_id
+        )
+        row_list = list(conn.execute(stat).fetchall())
+
+        # 如果`_journal_dt`不为空，则说明之前的日志中表明改股票数据已经部分完成，需要逐个核对日期
+        # 这里只把日期值不再`_journal_dt`的记录放入`row_list`
+        if _journal_dt is not None:
+            row_list = [row for row in row_list 
+                    if pd.to_datetime(row[1]) not in _journal_dt.index]
+            print(f"Resume job for {stock_id}, with {len(row_list)} rows left.")
+
+        num_rows = len(row_list)
+        # 如果行数为0，则说明是空数据，可以直接返回
+        if num_rows == 0:
+            return
+
+        if pbar:
+            #pbar.set_description(f"Did get the result set for stock {stock_id} from mssql")
+            pbar.set_description(f"Will work in paralle on dumping job on {stock_id} of len {num_rows}")
+        else:
+            print(f"Did get the result set for stock {stock_id} from mssql")
+
+        # 每一行是当个个股某一日的所有高频交易信息
+        # 使用多进程来加快速度
+        
+        #with Pool(self.num_workers if num_workers is None else num_workers) as pool:
+        if pool is None:
+            print("Will create new Pool object, but this is not encourage for large batch work.")
+            pool = Pool(self.num_worker)
+
+        # 在单个股票内部，对不同日期进行并行处理，对内存使用较为友好，不需要同时载入多个股票海量的全历史数据
+        with tqdm(total=num_rows, leave=False) as sub_pbar:
+            for _ in pool.imap_unordered(
+                functools.partial(
+                    DDBHFTLoader.dump_stock_daily_to_ddb, 
+                    type_name = type_name,
+                    stock_id = stock_id
+                ),
+                row_list
+            ):
+                sub_pbar.update()
 
         self.dump_journal_writer.write(f"{type_name},{stock_id},OK\n")
         self.dump_journal_writer.flush()