The kdb+ from Kx Systems is one of the fastest, most efficient and lightweight databases that has been used in high-frequency trading and other similar environments for almost two decades. Its popularity outside of finance has been much less pronounced, but nevertheless, it is arguably one of the most efficiently designed and optimized systems in the world of databases.

Kdb+ supports in-memory columnar storage from the outset and is technically an extension of the q programming language. A table in kdb+ is in essence a data structure in the q language. However, unlike similar concepts in other programming languages, a kdb+ table is enterprise-grade and can easily handle terabytes and petabytes of data.

Due to its inherent programming language, code that is written in q can be run against data stored in kdb+, so a custom function can be run in-database with very minimal effort.

Additionally, the size of the kdb+ binary is about 500 to 600KB, small enough to fit in the L3 cache of most modern CPUs.

Kdb+ also includes built-in MapReduce capabilities so that queries are automatically executed in parallel across muticore CPUs.

Business use case: Kdb+ is one of the most formidable databases to have existed in an enterprise setting. It was traditionally available only for perpetual core-based licensing, but in recent days, the company has added support for subscription-based and on-demand licensing. Its low footprint and simplicity of use makes it well suited for enterprise needs. However, this comes with a caveat. The q language is very terse and can appear cryptic to new users. The language, arguably, has a slightly steeper learning curve than others and requires practice and first-hand experience. That said, there are ample online resources to learn and utilize the features of the database. Native interfaces for R, Python, C, Java, and other programming languages, along with libraries used for machine learning, make it particularly well suited for data science workloads involving large datasets.