[Info-ingres] When is a table just to damn big?

[Karl and Betty Schendel]

> On Nov 5, 2018, at 10:20 AM, Martin Bowes <martin.bowes at ndph.ox.ac.uk> wrote:
> 
> Hi Karl,
> 
>> That's a weird message. 
> Yep.
> 
>> Do you need the table key to be btree?  Can it be hash? Hash partitioning doesn't necessarily play well with a btree index.
> We are doing lots of inserts into it and it's a big table so I suspect a hash structure would be nasty.

Lots of inserts into a hash isn't necessarily a bad thing, especially when the table
is large.  I did a fair amount of investigation on this subject back when Netsol ran
the DNS registry on Ingres.  As long as the keys hash reasonably well the overflow
chains average out to be something like 1.1 pages long for short records and
weekly modifications.

Karl