Is a UUID in Cassandra really "necessary?"

Question

I recently started using Cassandra - I come from a traditional relational database background, so it's definitely a bit different. One thing I'm used to always doing is generating a unique ID for each row (OID, etc.). So, for my tables that I've been creating in Cassandra I've been putting a UUID column on each of them and generating a UUID. My question is...is this really "necessary"? I'm not using the UUID as part of my partition key, so I'm not really using it for anything at the moment, but it's a tough habit to break. Some advice would be great!

Answer 1

Exactly it's not necessary. But introducing a UUID in a table may be useful in certain cases. For example imagine you have a table like :

CREATE TABLE user (
    id uuid,
    name text,
    login text,
    day_of_birth date
) PRIMARY KEY (login);

This table allows you to query users by login. Now imagine you also want to query users by name. Of course if this kind of query will be run just a few time, you can create a SECONDARY INDEX. But if you want to have good read performance, you can denormalize your data by having a table structure like :

CREATE TABLE user (
    id uuid,
    name text,
    login text,
    day_of_birth date
) PRIMARY KEY (id);

CREATE TABLE user_by_name (
    id uuid,
    name text
) PRIMARY KEY (name);

CREATE TABLE user_by_login (
    id uuid,
    login text
) PRIMARY KEY (login);

But with this structure, you have to insert and update in all 3 tables to maintain data. Instead of creating two other tables you can use MATERIALIZED VIEW to maintain only one table an let cassandra maintain view:

CREATE TABLE user (
    id uuid,
    name text,
    login text,
    day_of_birth date
) PRIMARY KEY (id);

CREATE MATERIALIZED VIEW user_by_name
AS 
SELECT *
FROM user
WHERE id IS NOT NULL
    AND name IS NOT NULL
PRIMARY KEY ((name), id);

CREATE MATERIALIZED VIEW user_by_login
AS
SELECT *
FROM user
WHERE id IS NOT NULL
    AND login IS NOT NULL
PRIMARY KEY ((login), id);