Tag
GreptimeDB explains its architecture of separating compute and storage via shared object storage in distributed deployments, and describes how it reuses SST data files during repartitioning to avoid migrating historical data wholesale, while also comparing the differences with ClickHouse.
Fly.io 介绍了 Sprites Block Device（SBD），用一个更简单的块设备方案取代了之前魔改的 JuiceFS 存储栈，以提升 Sprites 的可靠性并消除元数据不一致问题。
介绍 duckdb_object_storage——一个基于 SlateDB 与 Apache OpenDAL 的 DuckDB 社区扩展，可将原生 DuckDB 数据库直接存于 S3 等对象存储，无需额外运行目录服务或协调器，支持单写多读且性能接近本地原生 DuckDB。
一篇技术博客深入分析了把 Git 仓库放到对象存储(S3)上的常见误区，指出当前 S3-backed WAL 方案在写密集场景下的不足，并重新剖析了 Git 作为内容寻址存储加图数据库的数据模型对存储层的真实需求。
Infino explains how building a full-text and vector search index on object storage (S3, GCS, Azure Blob) shifts the key optimization target from CPU or bytes to minimizing network round trips, detailing its superfile/supertable architecture with manifests and term indexes.
Vectorize launches Hindsight 1.0, open-source memory for AI agents that learns from experience, alongside Hindsight Memlake, a Rust-based storage engine handling 100M memories at sub-100ms p99 without a database or consensus cluster.
VaultS3 是一个轻量级、单文件的 S3 兼容对象存储方案，闲置内存仅约 17MB，内置 Web 控制台，支持 AES-256 静态加密、版本控制、纠删码和集群复制，适合作为 MinIO 的轻量替代，用于小型 VPS 和边缘节点部署。
Delta is Meta's object storage service that uses chain replication to achieve high availability and strong consistency, designed for critical bootstrap and disaster recovery workloads.
RustFS 1.0 is released as an open-source, high-performance distributed object storage system built in Rust, offering a migration path for MinIO users after its maintenance ended.
This blog post explains how to run Git on object storage by inventing a new packfile format native to object storage, solving performance issues with traditional Git packfiles for production-sized repositories.
Tigris demonstrates using its object storage as a database substitute by implementing essential primitives such as unique constraints and transactions directly on its distributed key-value store built on FoundationDB.
Bookshelf is a self-hosted eBook library that uses object storage for managing and reading EPUB and PDF files, with support for Cloudflare R2 and local storage.
PicoMQ is a system for durable real-time streams over HTTP, built on S3-compatible object storage, allowing for independent and scalable streams per use case.
Walgit is a Git server implementation that runs as a single binary, using object storage like S3 or GCS as the backend, enabling scalable hosting of Git repositories without local state.
The article demonstrates a method for building fast drilldown dashboards from a single Parquet data cube using the Hyparquet JavaScript reader and HTTP range requests against object storage, eliminating the need for a traditional database.
This article explores storage engine write durability trade-offs, explaining how different storage layers (memory, local SSD, remote storage) move latency and durability, and analyzes the cost of operations in an append-only key-value store built on object storage.
Hugging Face introduces Storage Buckets, a scalable object storage service for AI teams with per-TB pricing, Xet deduplication, built-in CDN, and no git overhead, designed for datasets, model checkpoints, and ML artifacts.
A technical comparison between ZeroFS and Amazon S3 Files, two systems that expose POSIX filesystems backed by object storage. The article highlights differences in storage layout, object interoperability, and trade-offs between direct S3 access and ZeroFS's packing, compression, and encryption.
HydraDB OSS is an open-source graph database designed for AI applications, offering fast, low-latency performance with minimal operational overhead by leveraging object storage.
Firn is an open-source, multi-tenant vector and full-text search engine backed by object storage like AWS S3, providing a tiered storage architecture with RAM and NVMe caching for performance. It achieves sub-second cold queries with IVF_PQ indexes and microsecond warm hits via result caching.