Paras
Paras

Reputation: 642

Are there any performance penalties when using a TEXT as a Primary Key?

If yes, what would the data model look like if I want to have a unique TEXT field?

Upvotes: 8

Views: 2053

Answers (1)

Aaron
Aaron

Reputation: 57748

No. Regardless of data type used, Cassandra stores all data on disk (including primary key values) as hex byte arrays. In terms of performance, the datatype of the primary key really doesn't matter.

The only case where it would matter, is in token/node distribution. This is because the generated token for "12345" as text will be different from the token generated for 12345 as a bigint:

aploetz@cqlsh:stackoverflow> CREATE TABLE textaskey (key text PRIMARY KEY, value text);
aploetz@cqlsh:stackoverflow> CREATE TABLE longaskey (key bigint PRIMARY KEY, value text);
aploetz@cqlsh:stackoverflow> INSERT INTO textaskey (key, value) VALUES ('12345','12345');
aploetz@cqlsh:stackoverflow> INSERT INTO longaskey (key, value) VALUES (12345,'12345');
aploetz@cqlsh:stackoverflow> SELECT token(key),value FROM textaskey ;

 token(key)          | value
---------------------+-------
 2375712675693977547 | 12345

(1 rows)
aploetz@cqlsh:stackoverflow> SELECT token(key),value FROM longaskey;

 token(key)          | value
---------------------+-------
 3741197147323682197 | 12345

(1 rows)

But even in this example, one shouldn't perform faster/different than the other.

Upvotes: 9

Related Questions