Re: RMVM for SQL index objects

On 30-Jun-2016 06:34 -0500, Rob Berendt wrote:

There is a school of thought, disputed by some, that multiple
indexes aren't the beast that they used to be. Yes, I'm old enough
to remember numerous John Sears lectures on how he helped peoples
performance by reducing the very high (at the time) number of indexes
they had against their tables. Numerous "names" since then have said
that IBM has really changed stuff over the years that the performance
hit isn't nearly as severe.

Context matters greatly. Adding a few extra SMAPPed INDEXes in the context of general I/O might not be too big of an impact for the additional maintenance overhead. But for a dedicated reorganize, the number and size of the access paths over the data being reorganized can see a difference between having that data remain offline\unavailable for a couple time-units and having that data remain unavailable for possibly many|tens of time-units; those time-units still could be hours. So planning to avoid serialized rebuilds is a good idea; ensuring the rebuilds will be done in parallel after a large amount of data is reorganized is very sensible, whereas failing to make that happen can greatly impact the ability to access the data.