Simple question about partitioning

[Ken Camann]

I have a quick partitioning question. Suppose I make a table like this

CREATE TABLE t(
unique_id SERIAL,
group_id BIGINT NOT NULL,
data INT
);

and the table is huge. I know that most of the time, I only want rows that are in the same group, so I partition it by the groups (group_id). However, what happens when I want to look something up by the primary key (unique_id)? I assume that is very slow, i.e., the performance of the table being unpartitioned. Is that correct? And is there something else I should be doing here?

From what I understand, it will be possible to use foreign keys soon in partitioned table. Would that have the same performance characteristics as the above scenario? Although I would never SELECT rows with differing group_ids, it seems like a foreign key referencing the unique_id would not have the knowledge to take advantage of the partitioning system.

Thanks,
Ken