Designing the Data Model

Cassandra Query Language

Storage Type

Storge types document

Storage types Source code

Modeling by Query

table example

Relational Database

CREATE TABLE stock_symbol (
  symbol varchar PRIMARY KEY，
  description varchar,
  exchange varchar
);

CREATE TABLE stock_ticker (
  symbol varchar REFERENCES stock_symbol(symbol),
  tick_date varchar,
  open decimal,
  high decimal,
  close decimal,
  volume bigint,
  PRIMARY KEY (symbol, tick_date)
);

Select all the day close prices, and description of the stocks listed in the NASDAQ exchange on 24-APR-2014.

SELECT S.symbol, S.description, T.tick_date, T.close
  FROM stock_symbol S, stock_ticker T
  WHERE S.symbol = T.symbol
    AND S.exchange = 'NASDAQ'
    AND T.tick_date = '2014-04-24';

Cassandra database

There is no join operation.
To achieve the result, the answer is to use denormalization. It create a table which merge stock_symbol and stock_ticker together.

CREATE TABLE stock_ticker_by_exhange_date (
  exchange varchar,
  symbol varchar,
  description varchar,
  tick_date varchar,
  close decimal,
  PRIMARY KEY ((exchange, tick_date), symbol)
);

Important Considerations

  • Data dupliction
  • Bucketing
  • Valueless column
  • Time-series data

more details about how Cassandra handle time series data. http://www.datastax.com/dev/blog/advanced-time-series-with-cassandra

Indexing

row key determine the row location