Kolonniarvutused

Kolonniarvutused, also known as column-oriented databases or columnar databases, represent a different approach to storing and querying data compared to traditional row-oriented databases. In a row-oriented database, data is stored and retrieved row by row, meaning all values for a single record are kept together physically. This is efficient for transactional workloads where entire records are frequently accessed or modified.

In contrast, a columnar database stores data column by column. All values for a single column are stored together. This architecture is highly optimized for analytical workloads, such as data warehousing and business intelligence, where queries often involve aggregating or analyzing data across many rows but only for a subset of columns. For example, a query that calculates the average sales price across all transactions would only need to access the "sales price" column, and in a columnar database, it can retrieve all these values contiguously and quickly.

The advantages of this approach include significantly faster query performance for analytical tasks, better data compression ratios because data within a single column typically has similar characteristics, and reduced I/O operations for analytical queries. However, operations that involve inserting or updating entire rows can be less efficient compared to row-oriented systems. Popular examples of columnar databases include Amazon Redshift, Google BigQuery, Snowflake, and Apache Cassandra (which can be configured in a columnar fashion).