Database Reference
In-Depth Information
Is Cassandra Right for Me?
This isn't a very easy question to answer. Using Cassandra requires thinking in a
different way about how you store data. While there are rows and columns, Cas-
sandra is, at its base, a key/value store. There is no built-in full text search; there
are no B-tree indexes or data manipulation functions.
One of the biggest differences between Cassandra and standard SQL RDBMSs
is that there are no manipulation functions. These include
SUM, GROUP,
JOIN, MAX, MIN,
and any other method you would use to modify the data at
query time.
While deciding if Cassandra is a good fit for your use case, know that a lot
of data manipulation can be achieved at write-time rather than read-time. This, of
course, means that you will be storing different views of the same data in multiple
places. This is not necessarily a bad thing.
One example of this is to use counter columns where you would need aggreg-
ation. This is as easy as incrementing a value for each of the different ways you
want to see your data. This pattern does require that you know what questions you
want to ask ahead of time; if you need ad hoc data analysis in real time, Cassandra
may not be the right fit.
Cassandra Terminology
In order to understand Cassandra, a good place to start is the vocabulary.
Cluster
A cluster is two or more Cassandra instances working together. These instances
communicate with each other using the Gossip protocol.
Homogeneous Environment
Cassandra is considered homogeneous. This means that each and every Cassandra
node contains everything required to complete a cluster. This differs from other
systems such as HBase, which have master servers, region servers, ZooKeeper
servers, and other different types of nodes. With Cassandra, expanding a cluster
is as easy as adding a new node that is identical to every other node with the ex-
ception of the configuration and data for which it is responsible. This takes some
complexity out of managing an infrastructure with many nodes.
