Hi David,

The insert into a heap should be very fast - no matter how big it
is. But the thing that could slow it down is a secondary index on the
table. The progressivly slower insert smacks of an ISAM secondary
index in overflow. The overflows on the ISAM would require
sequentially searching every overflow page. Hence the performance
you observed.

The subsequent modify to btree may have dropped the index, it
may have cleaned up the index. Given that the subsequent inserts
proceeded much faster and didn't degrade - my guess is that the index
was not persistent.

Martin Bowes
--
Random Farscape Quote #16:
John - I'd like to leave this planet the way I got here. Alive and single.