Postgres on K8s with fast data branching

Xata는 Kubernetes 위에서 다수의 Postgres 인스턴스를 자체 호스팅할 수 있는 오픈소스 클라우드 네이티브 플랫폼으로, 저장소 수준의 Copy-on-Write 기반 빠른 브랜칭과 자동 스케일-투-제로 기능을 제공한다. 고가용성, 자동 장애 조치, 무중단 업그레이드, 스토리지와 컴퓨트 분리, 서버리스 SQL 드라이버 등 운영 편의성을 갖추었으며, 대규모 데이터 복사와 비용 효율적인 개발/테스트 환경 구축에 적합하다. CloudNativePG와 OpenEBS를 기반으로 하며, REST API와 CLI를 통해 제어가 가능하다. 단일 인스턴스 용도나 공개 PGaaS 제공에는 적합하지 않다.

https://github.com/xataio/xata

#postgresql #kubernetes #cloudnative #copyonwrite #scaletozero

Copy‑on‑write (COW) là kỹ thuật quan trọng giúp Swift thực hiện semantics giá trị một cách hiệu quả, giảm sao chép dữ liệu không cần thiết và tối ưu bộ nhớ. Bài viết phân tích sâu cách COW được áp dụng trong Swift internals, từ cấu trúc dữ liệu tới quản lý bộ nhớ, hữu ích cho lập trình viên muốn hiểu cơ chế hoạt động bên trong ngôn ngữ. #Swift #Programming #CopyOnWrite #LậpTrình #SwiftInternals #iOS

https://www.reddit.com/r/programming/comments/1qh97lb/copyonwrite_teaches_you_everything_about_s

Что такое снапшоты в СХД и как мы их реализовали в TATLIN.BACKUP

Еще год назад мы считали, что снапшоты без репликации не являются приоритетом для нашей команды. Но практика показала обратное: вирусы-шифровальщики, сбои и ошибки могут повредить резервные копии и нарушить их целостность. Чтобы защитить данные и дать возможность быстро восстановить их в неизменном виде, мы реализовали поддержку снапшотов в TATLIN.BACKUP. Новый функционал уже доказал свою эффективность и получил положительные отзывы клиентов. Меня зовут Ростислав, я эксперт по разработке ПО отдела систем обработки данных в YADRO. Сегодня разберемся, как устроены наши снапшоты и какие у них особенности. Для этого совершим экскурс в мир систем хранения данных (СХД) и рассмотрим их устройство.

https://habr.com/ru/companies/yadro/articles/941358/

#снапшоты #TATLINBACKUP #TBoost_SDK #RedirectonWrite #CopyonWrite #MarkandSweep #reference_counting #СХД

Oof, i forgot to create a swap file!

I usually like to have a small swap space, so the dreaded OOM-Killer doesn't sacrifice my children. 😅

That's gonna conflict with my btrfs compression plan, IIRC. (i'm using btrfs, because it supports compression) No worries, there are work-arounds.

Correction: syslog says it conflicts with copy-on-write

#Linux #btrfs #CopyOnWrite #swap

#Swift uses something like this:

“Simple Fast Value Semantics”, Tibor ‘zupa’ Halter, #Boomla (https://boomla.com/docs/simple-fast-value-semantics).

“isKnownUniquelyReferenced(_:)” (https://developer.apple.com/documentation/swift/isknownuniquelyreferenced(_:)-5kvtu).

Via Lobsters: https://lobste.rs/s/xo7a4v/simple_fast_value_semantics

On HN: https://news.ycombinator.com/item?id=41138550

#PLDI #CopyOnWrite

Copy-on-write on APFS

APFS (like many modern file systems but unlike its predecessor HFS+) supports copy-on-write. This means you can logically copy a file - it looks and behaves like a distinct file - but it doesn't immediately copy the file's contents on disk - it merely shares them with the original. Only if and as you modify either version do they start to diverge on disk, with APFS dynamically allocating new storage for the modified parts1.

This is kind of a sister function to hard links, which similarly [...]

https://wadetregaskis.com/copy-on-write-on-apfs/

Utilities for Copy-On-Write filesystems, call that cowtools.

#cowtools #CopyOnWrite #filesystems #QCOW2 #Btrfs (both are COW so I figure it would be relevant to anyone using them)

Should #coreutils #dd support CoW / reflink / FICLONE / copy_file_range optimizations?

Currently, cp has support for this, but can only copy whole files.
If I want to combine sections of files into an output file, there is no simple userspace tool to do so with #CopyOnWrite semantics.
#linux #filesystem