Help performance tune insert as select of lobs

Question

I am using Oracle 11.2.0.3 Enterprise edition.
I am looking at means to improve following insert. Any suggestion would be highly appreciated.
Below is tkprof report. tab1 is created as nologging. The insert is geting blob(book_file) over a db link. Lob column is also declared nocache nologging.
Also please help me understand the tkprof report. As per the report elasped time - cpu time = 12.31 -2.90 = 9.41. However, sum of wait evens is well over 9.41s.
Thanks for you time.
INSERT /*+ APPEND */ INTO tab1
     SELECT   B.book_file_ID,
         bv.file_NB,
         bv.archvd_fl,
         bv.flng_pk,
         bv.book_id,
         bv.x_insrt_dtm,
         bv.x_updt_dtm,
         BV.book_FILE
  FROM   tab1_MV B, tab1@db_link1 BV
WHERE   B.book_file_ID = BV.book_file_ID
         AND NOT EXISTS (SELECT   1
                           FROM   tab1 i
                          WHERE   I.book_file_ID = B.book_file_ID)
call     count       cpu    elapsed       disk      query    current        rows
Parse        1      0.02       0.68          0          7          1           0
Execute      1      2.88      11.63       1342      82297      49123          99
Fetch        0      0.00       0.00          0          0          0           0
total        2      2.90      12.31       1342      82304      49124          99
Misses in library cache during parse: 1
Optimizer mode: ALL_ROWS
Parsing user id: 49 
Number of plan statistics captured: 1
Rows (1st) Rows (avg) Rows (max)  Row Source Operation
         0          0          0  LOAD AS SELECT  (cr=82297 pr=1342 pw=2381 time=11632815 us)
        99         99         99   FILTER  (cr=60811 pr=86 pw=0 time=24355 us)
     60240      60240      60240    MERGE JOIN  (cr=50 pr=0 pw=0 time=582671 us cost=325 size=10963680 card=60240)
    102207     102207     102207     REMOTE  tab1 (cr=0 pr=0 pw=0 time=477887 us cost=186 size=18079665 card=102145)
     60240      60240      60240     SORT JOIN (cr=50 pr=0 pw=0 time=72327 us cost=139 size=301200 card=60240)
     60240      60240      60240      MAT_VIEW ACCESS FULL tab1_MV (cr=50 pr=0 pw=0 time=7201 us cost=21 size=301200 card=60240)
     60141      60141      60141    INDEX UNIQUE SCAN PKCBV_book_file_ID (cr=60761 pr=86 pw=0 time=495753 us cost=1 size=5 card=1)(object id 1634074)
Elapsed times include waiting on following events:
  Event waited on                             Times   Max. Wait  Total Waited
  ----------------------------------------   Waited  ----------  ------------
  single-task message                             1        0.05          0.05
  SQL*Net message from dblink                   181        0.07          0.49
  SQL*Net message to dblink                     179        0.00          0.00
  SQL*Net more data from dblink                6878        0.01          0.19
  db file sequential read                      1342        0.03          1.57
  log file sync                                   1        0.02          0.02
  Disk file operations I/O                        9        0.00          0.00
  direct path write                            1790        0.01          0.19
  direct path sync                                5        0.01          0.05
  control file sequential read                 7337        0.00          0.07
  control file parallel write                  4888        0.01          7.48
  asynch descriptor resize                     1492        0.00          0.00
  enq: CF - contention                            1        0.01          0.01
  SQL*Net message to client                       1        0.00          0.00
  SQL*Net message from client                     1        6.86          6.86
SQL ID: gb9krpvpav4p5 Plan Hash: 0

Answer 1

spur230 wrote:
Tubby wrote:
Since the table with the blob you need is on a remote database you may be best served by processing the query on the remote site instead of the local one.
http://jonathanlewis.wordpress.com/2010/10/07/distributed-pipelines/
Would be something i'd look in to implementing for this.
Cheers,Thanks Tubby.
I did read about this option . However, I found driving_site hint is not applicable for DML operations.
http://jonathanlewis.wordpress.com/2008/12/05/distributed-dml/
A distributed DML statement must execute on the database where the DML target resides. The DRIVING_SITE hint cannot override this.
True, unless you slap it inside a pipelined function as demonstrated in the link i sent (which actually links to the link you are quoting from) :)
One other thing, i didn't understand the purpose of joining to tab1_MV B in your query. Just another existence test?
Cheers,