Any way to reduce waits over the link? (SQL*Net message from dblink)

The queries are fine (not pulling whole tables) The network is fine (running at about half the capacity) Db's are not direct-connected, traffic goes over a switch, 100Mbps ethernet.
When looking at trace files, it looks like oracle does some kind of checking over the link to make sure the objects are valid, that might be a contributor to these waits, any way to get rid of this? Some queries have already been 'tuned' to use snapshots, but some other queries require instanteneous updates to local tables, so snapshots won't work.

My last statspack for an hour shows:

                                                           Avg
                                      Time TotalWait wait  Waits
Event                           Waits outs Time(cs)  (ms)   /txn
--------------------------- -------- ---- -------- ---- ------
SQL*Net message from dblink 1,428,648   55  454,719    3    7.1
db file sequential read       588,295    0  440,978    7    2.9
log file sync                  94,051    5   67,230    7    0.5
db file parallel write         31,395    0   44,485   14    0.2

454,719cs/100/60=75mins of wait time on a 10cpu machine the file reads cannot really be tuned, unless i increase the buffer cache a bit, but it's running at 99% hit rate. log file sync's can't be tuned either, and they don't bother me much, redo is running on dedicated disks, striping didn't help. .......
We use Oracle 8.1.7.4 on Solaris 2.7 boxes remove NSPAM to email Received on Wed Sep 10 2003 - 11:41:19 CDT