CODEWITHSUNDEEP
cassandrastorage
user2984857
user2984857

Reputation: 3

Cassandra Storage Limit

I am trying to understand Cassandra limitation as our Cassandra DB grows. Reading from "http://wiki.apache.org/cassandra/CassandraLimitations", it says The maximum number of cells (rows x columns) in a single partition is 2 billion. . While "http://www.pcworld.idg.com.au/article/373483/new_cassandra_can_pack_two_billion_columns_into_row/" says it can pack 2 billion columns into a row. My understanding is that a row will not be split into different partitions. Therefore this means a partition has to be able to hold more than 2 billion cells, which is contrast to the first post. So what is the true limit on this?

Upvotes: 0

Views: 942

Answers (1)

Aaronontheweb
Aaronontheweb

Reputation: 8414

These articles are confusing because they're using two different names for the same thing.

Any given row in a Cassandra column family can store up to 2 billion columns, and the next row in the same column family could have another 2 billion distinctly different columns than the preceding row.

Wide rows are definitely split into separate SSTable files on-disk, but they are not partitioned across the network. Columns are physically sorted (by name) and stored next to each other on disk.

The purpose of Cassandra's compaction process is to group SSTable files for rows and column families into as few files as possible in order to maintain maximum performance, and there are different compaction options that provide better performance for read-heavy vs. write-heavy (default) scenarios.

Upvotes: 1

Related Questions