Dynamic Fields vs. Multicore

[Briggs Thompson <w....@gmail.com>]

Hi All,

I was searching around for documentation of the performance differences of
having a sharded, single schema, dynamic field set up vs. a multi-core,
static multi-schema setup (which I currently have), but I have not had much
luck finding what I am looking for. I understand commits and optimizes will
be more intensive in a single core since there is more data (though I would
offset by sharding heavily), but I am particularly curious about the search
performance implications.

I am interested in moving to the dynamic field setup in order to implement a
better global search, but I want to make sure I understood the drawbacks of
hitting those datasets individually and globally after they are merged
(NOTE: I would have a global field signifying the dataset type, which could
then be added to the filter query in order to create the subset for
individual dataset queries).

Some background about the data: it is extremely variable. Some documents
contain only 2 or 3 sentences, and some are 20 page extracted PDFs. There
would probably only be about 100-150 unique fields.

Any input is greatly appreciated!

Thanks,
Briggs Thompson