Hive HQL - optimizing WINDOW function

I have following HQL excuted by MR engine where the source table has almost 800 million records

select concat(upp_sys_id,'#',min(bhv_tm) over ssn,'#',ssn_seq_all) as ssn_id
,evt_drt
,row_number() over ssn as ssn_seq
,`dw_dat_dt` ,`msg_id` ,`evt_nm` ,`upp_sys_id`
,`bhv_tm`
from test.VT_seq_all
window ssn as (partition by upp_sys_id,ssn_seq_all order by bhv_tm);

most of the reducers take about 10min to finish, but there is a reducer has running for 4 hours. The input records of the abnormal task are similar to others. In general, the number of records with same upp_sys_id and ssn_seq_all is less than 1 thousands. But for a few specific upp_sys_id and ssn_seq_all, the number of records is more than 100 thousands.

Is there any way to optimize this hql?

Here is the execution plan

STAGE DEPENDENCIES:
  Stage-1 is a root stage
  Stage-0 depends on stages: Stage-1

STAGE PLANS:
  Stage: Stage-1
    Map Reduce
      Map Operator Tree:
          TableScan
            alias: VT_seq_all
            Statistics: Num rows: 782125303 Data size: 2610606354398 Basic stats: COMPLETE Column stats: NONE
            Reduce Output Operator
              key expressions: upp_sys_id (type: string), ssn_seq_all (type: bigint), bhv_tm (type: string)
              sort order: +++
              Map-reduce partition columns: upp_sys_id (type: string), ssn_seq_all (type: bigint)
              Statistics: Num rows: 782125303 Data size: 2610606354398 Basic stats: COMPLETE Column stats: NONE
              value expressions: evt_drt (type: bigint), dw_dat_dt (type: date), msg_id (type: string), evt_nm (type: string)
      Reduce Operator Tree:
        Select Operator
          expressions: VALUE._col0 (type: bigint), KEY.reducesinkkey1 (type: bigint), VALUE._col1 (type: date), VALUE._col2 (type: string), VALUE._col3 (type: string), KEY.reducesinkkey0 (type: string), KEY.reducesinkkey2 (type: string)
          outputColumnNames: _col0, _col1, _col2, _col3, _col4, _col5, _col9
          Statistics: Num rows: 782125303 Data size: 2610606354398 Basic stats: COMPLETE Column stats: NONE
          PTF Operator
            Function definitions:
                Input definition
                  input alias: ptf_0
                  output shape: _col0: bigint, _col1: bigint, _col2: date, _col3: string, _col4: string, _col5: string, _col9: string
                  type: WINDOWING
                Windowing table definition
                  input alias: ptf_1
                  name: windowingtablefunction
                  order by: _col9 ASC NULLS FIRST
                  partition by: _col5, _col1
                  raw input shape:
                  window functions:
                      window function definition
                        alias: min_window_0
                        arguments: _col9
                        name: min
                        window function: GenericUDAFMinEvaluator
                        window frame: RANGE PRECEDING(MAX)~CURRENT
                      window function definition
                        alias: row_number_window_1
                        name: row_number
                        window function: GenericUDAFRowNumberEvaluator
                        window frame: ROWS PRECEDING(MAX)~FOLLOWING(MAX)
                        isPivotResult: true
            Statistics: Num rows: 782125303 Data size: 2610606354398 Basic stats: COMPLETE Column stats: NONE
            Select Operator
              expressions: concat(_col5, '#', min_window_0, '#', _col1) (type: string), _col0 (type: bigint), row_number_window_1 (type: int), _col2 (type: date), _col3 (type: string), _col4 (type: string), _col5 (type: string), _col9 (type: string)
              outputColumnNames: _col0, _col1, _col2, _col3, _col4, _col5, _col6, _col7
              Statistics: Num rows: 782125303 Data size: 2610606354398 Basic stats: COMPLETE Column stats: NONE
              File Output Operator
                compressed: false
                Statistics: Num rows: 782125303 Data size: 2610606354398 Basic stats: COMPLETE Column stats: NONE
                table:
                    input format: org.apache.hadoop.mapred.SequenceFileInputFormat
                    output format: org.apache.hadoop.hive.ql.io.HiveSequenceFileOutputFormat
                    serde: org.apache.hadoop.hive.serde2.lazy.LazySimpleSerDe
            Stage: Stage-0
              Fetch Operator
                limit: -1
                Processor Tree:
                  ListSink