Conference ulysse::rdb_vms_competition

383.0. "RDB/ORACLE PERFORMANCE?" by ALBANY::MULLER (Fred Muller) Sun Jul 02 1989 04:57

    I do not get into these conferences very often, but I recall reading
    here or elsewhere about the games that Oracle (and maybe other
    competitors) play to upgrade performance by putting everything in
    memory. 
    
    I am just doing a multi-file database job (2 dozen areas/files)
    RDB030A/DTR042/VMS047/CDD+V?/=VAX3600/RA82(nocluster) job and because
    the cpu is one of the first systems on which I have had the luxury of
    lots of memory, I chose to use make a change from the default of 20 RDB
    buffers to 100 RDB buffers. 
    
    Guess shat?  In retrospect, I guess the expected happened, as it should
    have - just like Oracle. The high I/O disk rate dropped to next to
    nothing and the CPU usage went to nearly 100% on a dedicated system.
    
    Imagine that!  Sales/sales-support/bench-markers take note. Anyone else
    out there as thick as me - or is there another explanation?  Anyone
    have any reservations about the RDB recovery process in this situation?
    I picked 500 recovery buffers too.
    
    I am cross posting this in both the RDB_VMS_COMPETITION and RDB
    conferences. 
    
    Fred, SWS/PSS (@ALO)