September 2016 Summaries
2 posts from Cockroach Labs
Filter
Month:
Year:
Post Summaries
Back to Blog
CockroachDB, a distributed SQL database built upon a transactional key-value store, has improved its performance by implementing column families, a technique commonly used in NoSQL databases. Initially, CockroachDB's format version stored each non-primary index column as a separate key-value entry, leading to inefficiencies such as repetitive data and transaction-level overheads. By grouping multiple columns into a single key-value entry, column families have allowed for significant performance enhancements, especially for tables with many columns. This change has been particularly successful, with benchmarks showing a fivefold increase in speed for operations on tables with 20 integer columns. The implementation also maintains backward compatibility and allows users to manually optimize column family assignments based on their query patterns. These advancements bring CockroachDB closer to its 1.0 release, showcasing a blend of SQL and NoSQL optimizations.
Sep 29, 2016
1,135 words in the original blog post.
CockroachDB, known for its scalability across multiple nodes and reliability even in the face of hardware failures, was tested on a Raspberry Pi to explore its capabilities on the ARM architecture, despite its typical use on x86 systems. The team successfully ran CockroachDB on a Raspberry Pi Model B, an ARMv6 architecture with 512 MB of memory, using Raspbian Jessie as the operating system. The process involved compiling the database from source, adjusting constants and compiler flags for compatibility, and managing memory constraints by altering the Raspberry Pi's memory configuration. Despite achieving a functional database, the experiment comes with cautionary notes regarding potential bugs and risks, as CockroachDB does not officially support ARM or 32-bit architectures. The endeavor is shared with the community, encouraging feedback and further exploration, and inviting users to report issues via GitHub.
Sep 20, 2016
1,136 words in the original blog post.