[gdulli]

1. Flexibility/freedom to distribute partitions in the future if needed.

2. Indexing doesn't work well in all cases. You can be better off scanning entire small partition tables lacking an index on a given column than with a single very large table whether that column has an index or not. (Indexes take up space and need to be read from disk if they don't fit in a memory cache, indexes don't work well for low-cardinality columns, etc.)

3. There are operations you can parallelize on a large number of small/medium tables and perform faster or more conveniently than a single very large table. One of my favorite techniques:

# usage: seq 0 255 |parallel -j16 ./alter-tablespace.sh {}

hexn=`printf "%02x" $1`

psql -Atc "ALTER TABLE tablename_${hexn} SET TABLESPACE new_tblspace" -hxxx -Uxxx xxx

4. A nice side effect of properly/evenly partitioned data you get for free is that you can do certain types of analysis on a single partition (or a few) very quickly and have it represent a sampled version of the data set. You can think of it as another index you get for free.