I have a database that supports a web application with several large tables. I'm wondering if partitioned tables will help speed up certain queries. Each of these tables has a colum called client_id. Data for each client_id is independent from every other client_id. In other words, web queries will always contain a where clause with a single client_id. I'm thinking this may be a good column on which to partition my large tables.
After reading up on partitioned tables, I'm still a little unsure as to how best to partition. For example, a typical table may have 50 million rows distributed more or less evenly across 35 client_ids. We add new client_ids periodically but in the short term the number of client_ids is relatively fixed.
I was thinking something along these lines:
CREATE TABLE foo (
id INT NOT NULL PRIMARY KEY,
... more column defs here...
client_id int
)
PARTITION BY KEY(client_id)
PARTITIONS 35;
My question. Is this an optimal strategy for partitioning these types of tables? My tests indicate a considerable speedup over indexing on client_id, but can I do better with some other form of partitioning (i.e. hash or range)?