Conference orarep::nomahs::dec_data_distributor

244.0. "Troubleshooting needed" by ORAREP::HGOVC::PATRICKNG () Mon Oct 14 1996 00:55

    Hi,
    
    A customer has tested Oracle Data Replicator (fomerly called Digital
    Data Distributor) on an AlphaServer 2100 which is running OVMS 6.1 and
    RDB 6.0 with 1GB memory.
    
    The performance is very slow and the customer would like to find out
    the problems. Unfortunately, local Oracle technical support does not
    have expertise on this products. We likes to help Oracle to help the
    problem and leave Orale to resolve it for the customer.
    
    Symtoms:
    
    The process slowed down after proceeded 14 hours (7 million records has
    been replicated out of 20 millions). The customer does not accept 20
    times slower as compared when it was started at the beginning.
    
    The replication was coping 12DB RDB within the AlphaServer 2100. The
    system was dedicated for this testing. No network is involved.
    
    Could someone give us some advices to do the troubleshooting?
    
    
    Thanks,
    Patrick
    
    
    

    
    Patrick,
    
    It sounds as if we might need someone to work with the customer for
    awhile to do some experimenting and digging to find out what is going
    on.  I'm guessing there is some sort of thrashing, and perhaps some
    system or process quotas need to change.  Perhaps buffers need to be
    made larger.  What evidence does the customer have that there is a big
    slowdown?
    
    I think it is unlikely that we'ld get to the bottom of this just by
    posting a note or two and replies, so your customer should submit a
    problem report on which we can followp up.

If they are trying to do 20,000,000 rows in a single read write 
transaction, then they are doubtlessley seeing a huge amount of page 
faulting due to the number of record locks (or in fact RPLL's).

Cheers,

Mark.