A journey toward the columnar data store - lessons from past discussions and new proposals скачать в хорошем качестве

A journey toward the columnar data store - lessons from past discussions and new proposals

Presented by Aya Iwata at PGConf.dev 2025 (https://2025.pgconf.dev) Vertical Clustered Index (VCI) is an in-memory columnar index which provides easy access to column data. This feature enables faster scanning and aggregating large volumes of data. In addition, it has features to overcome the difficulty of updating and deleting data, which is a weakness of typical columnar storage. From 2016 to 2017, we proposed this feature which was implemented with index access methods and modification of the PostgreSQL core code to the community. However, we were unable to commit to it due to the following points. Foreign data wrapper or table access method are prefered for the implementation of columnar storage. Any development of a columnar storarge should be done with caution and avoid unnecessary restrictions on the PostgreSQL core code. This time, we decided to reflect past suggestions and make a new proposal to the community, as we thought this function would be useful for users. This talk covers: Introduction - Columnar storage and VCI. Summary of past discussions on the -hackers The structure and design decisions of the proposed VCI (Dependent on table access method). Performance comparison between PostgreSQL with VCI and without VCI. https://www.pgevents.ca/events/pgconf... 🎬 See more PGConf.dev 2025 videos at    • PGConf.dev 2025   Connect with us: LinkedIn:   / post   X   / pgconfdev   Mastodon: https://mastodon.social/@pgconfdev Bluesky: https://bsky.app/profile/pgconf.dev Web: https://pgconf.dev #postgresql #PGConfDev