diff --git a/src/client/src/tscGlobalmerge.c b/src/client/src/tscGlobalmerge.c
index f84ed0af9ba0979dad42b17958a715790214df13..6974e8e868021faf89bbbd1da2e39235aaa67d3c 100644
--- a/src/client/src/tscGlobalmerge.c
+++ b/src/client/src/tscGlobalmerge.c
@@ -233,7 +233,7 @@ static int32_t tscFlushTmpBufferImpl(tExtMemBuffer *pMemoryBuf, tOrderDescriptor
 
   // sort before flush to disk, the data must be consecutively put on tFilePage.
   if (pDesc->orderInfo.numOfCols > 0) {
-    tColDataSort(pDesc, (int32_t)pPage->num, 0, (int32_t)pPage->num - 1, pPage->data, orderType, true);
+    tColDataMergeSort(pDesc, (int32_t)pPage->num, 0, (int32_t)pPage->num - 1, pPage->data, orderType);
   }
 
 #ifdef _DEBUG_VIEW
diff --git a/src/query/inc/qExtbuffer.h b/src/query/inc/qExtbuffer.h
index bce90ae732a0501f8525fa6826d4213e801cb7e8..ceda7f98b3981b174a275467438b72d7541d14f7 100644
--- a/src/query/inc/qExtbuffer.h
+++ b/src/query/inc/qExtbuffer.h
@@ -227,7 +227,10 @@ void tColModelAppend(SColumnModel *dstModel, tFilePage *dstPage, void *srcData,
 
 typedef int (*__col_compar_fn_t)(tOrderDescriptor *, int32_t numOfRows, int32_t idx1, int32_t idx2, char *data);
 
-void tColDataSort(tOrderDescriptor *, int32_t numOfRows, int32_t start, int32_t end, char *data, int32_t orderType, bool stableSort);
+void tColDataQSort(tOrderDescriptor *, int32_t numOfRows, int32_t start, int32_t end, char *data, int32_t orderType);
+
+void tColDataMergeSort(tOrderDescriptor *, int32_t numOfRows, int32_t start, int32_t end, char *data, int32_t orderType);
+
 
 void taoscQSort(void** pCols, SSchema* pSchema, int32_t numOfCols, int32_t numOfRows, int32_t index, __compar_fn_t compareFn);
 
diff --git a/src/query/src/qExtbuffer.c b/src/query/src/qExtbuffer.c
index b8ca3f18b47e4edb3269b81c974c775f59537601..a004e0a60da9cdfd90ec450a9fba9d83b159a625 100644
--- a/src/query/src/qExtbuffer.c
+++ b/src/query/src/qExtbuffer.c
@@ -825,7 +825,7 @@ static void columnwiseQSortImpl(tOrderDescriptor *pDescriptor, int32_t numOfRows
   }
 }
 
-void tColDataSort(tOrderDescriptor *pDescriptor, int32_t numOfRows, int32_t start, int32_t end, char *data, int32_t orderType, bool stableSort) {
+void tColDataQSort(tOrderDescriptor *pDescriptor, int32_t numOfRows, int32_t start, int32_t end, char *data, int32_t orderType) {
   // short array sort, incur another sort procedure instead of quick sort process
   __col_compar_fn_t compareFn = (orderType == TSDB_ORDER_ASC) ? compare_sa : compare_sd;
 
@@ -844,8 +844,6 @@ void tColDataSort(tOrderDescriptor *pDescriptor, int32_t numOfRows, int32_t star
 
   if (end - start + 1 <= 8) {
     tColDataInsertSort(pDescriptor, numOfRows, start, end, data, compareFn, buf);
-  } else if (stableSort) {
-    columnwiseMergeSortImpl(pDescriptor, numOfRows, start, end, data, orderType, compareFn);
   } else {
     columnwiseQSortImpl(pDescriptor, numOfRows, start, end, data, orderType, compareFn, buf);
   }
@@ -853,6 +851,34 @@ void tColDataSort(tOrderDescriptor *pDescriptor, int32_t numOfRows, int32_t star
   free(buf);
 }
 
+void tColDataMergeSort(tOrderDescriptor *pDescriptor, int32_t numOfRows, int32_t start, int32_t end, char *data, int32_t orderType) {
+  // short array sort, incur another sort procedure instead of quick sort process
+  __col_compar_fn_t compareFn = (orderType == TSDB_ORDER_ASC) ? compare_sa : compare_sd;
+
+  SColumnModel* pModel = pDescriptor->pColumnModel;
+
+  size_t width = 0;
+  for(int32_t i = 0; i < pModel->numOfCols; ++i) {
+    SSchema* pSchema = &pModel->pFields[i].field;
+    if (width < pSchema->bytes) {
+      width = pSchema->bytes;
+    }
+  }
+
+  char* buf = malloc(width);
+  assert(width > 0 && buf != NULL);
+
+  if (end - start + 1 <= 8) {
+    tColDataInsertSort(pDescriptor, numOfRows, start, end, data, compareFn, buf);
+  } else {
+    columnwiseMergeSortImpl(pDescriptor, numOfRows, start, end, data, orderType, compareFn);
+  }
+
+  free(buf);
+}
+
+
+
 /*
  * deep copy of sschema
  */