This article explains the core concepts and architectures of distributed storage, covering file‑based systems like HDFS, object storage such as Ceph, block storage for high‑performance workloads, and key‑value stores like Redis and Cassandra, highlighting their use cases and design principles.
This article explains the concepts, advantages, and disadvantages of block storage, file storage, and object storage, compares their architectures, and clarifies when each type is appropriate for different applications and workloads.
This article explains how a gRPC bidirectional streaming read method was added to Ozone to reduce chunk‑by‑chunk request gaps, describes the client‑side implementation, presents single‑ and multi‑threaded performance tests showing roughly 30% faster reads, and outlines future enhancements such as pre‑fetching.
Block, file, and object storage each offer distinct architectures and trade‑offs—block storage provides high performance but limited sharing, file storage enables easy sharing at lower speeds, and object storage combines both advantages with metadata‑driven distribution—this guide explains their principles, benefits, drawbacks, and ideal applications.
Curve is a high‑performance, easy‑to‑operate, cloud‑native open‑source distributed storage system (CNCF Sandbox) that provides block and file storage for OpenStack, Kubernetes, and PolarFS, featuring Raft‑based consistency, hybrid storage, high availability, and an ongoing roadmap for AI and other workloads.
This article presents Baidu Cloud's block storage architecture, comparing replication and erasure‑coding fault‑tolerance methods, detailing the challenges of applying EC to mutable block data, and describing a two‑layer append‑engine solution with selective 3‑replica caching, cost‑benefit compaction, and performance optimizations for low‑cost, high‑throughput storage.
This article explains Baidu Intelligent Cloud's block storage (CDS) architecture, comparing fault‑tolerance methods, detailing the challenges of large‑scale erasure‑coded storage, and describing Baidu's two‑layer append‑engine solution that reduces I/O amplification while keeping costs low.
This article explains common storage types—file (NAS), block (EBS), and object (OSS)—detailing their physical foundations, performance metrics, underlying Alibaba Cloud architectures, protocol differences, and ideal application scenarios to help readers select the most suitable storage solution for any workload.
The forum gathers senior experts from Baidu, Meituan, Ant Group, Xiaomi and Baidu Cloud to share practical experiences and future trends on constructing ultra‑large‑scale file, block, KV and NoSQL storage systems, focusing on low‑cost, high‑performance solutions and architectural challenges.
As global data volumes surge toward 163 zettabytes by 2025, traditional file and block storage struggle to scale, making object storage—offering unique IDs, metadata binding, and effortless scalability—the preferred solution for handling massive, mostly unstructured data in modern cloud environments.
The article explains the core concepts, advantages, and limitations of file storage, block storage, and object storage, illustrating how each format organizes data, interacts with operating systems, and fits various application scenarios such as databases, cloud services, and unstructured data handling.
In this interview, NetEase’s Curve project lead Wang Pan explains the architecture, block and file storage services, performance advantages over Ceph, management modules, ongoing CPU and I/O optimizations, deployment tooling, and the open‑source roadmap that positions Curve as a high‑performance, cloud‑native storage solution.
Facing Ceph’s high latency, jitter, and upgrade constraints, NetEase Cloud Music adopted Curve block storage, which delivers over twice the IOPS, sub‑2‑second upgrade pauses, and stable, low‑latency I/O across 40 GB‑4 TB volumes, enabling the platform to meet its 99.99% availability SLA for billions of users while supporting future cloud‑native services.
This article outlines the historical development of storage systems—from single-node and centralized architectures to distributed, cloud, and cloud‑native solutions—detailing storage types (block, file, object), data models, system components, and key design principles for modern storage infrastructures.
This article provides a detailed overview of the evolution of storage systems—from single‑node to cloud‑native—covering fundamental concepts, data classifications, storage types such as block, file, and object, and examples like Rook and MinIO, aimed at beginners seeking a macro understanding of storage architecture.
This article explains the concepts of file, object, and block storage, compares their architectures and access methods, and outlines practical scenarios where each storage type best fits, helping you choose the optimal solution for your workload.
The article explains the rapid growth of global data, compares traditional file and block storage with modern object storage, and highlights why object storage is better suited for massive, unstructured data and cloud scalability.
This article outlines the evolution of storage systems—from early single‑node solutions to modern cloud‑native architectures—detailing key storage types (block, file, object), models, and design principles of centralized, distributed, and cloud storage, providing a concise roadmap for newcomers.
This guide explains the differences between block, file, and object storage, outlines their advantages and drawbacks, introduces distributed storage concepts, and provides a detailed, code‑rich walkthrough for deploying a Ceph cluster—including environment setup, node configuration, CRUSH map tuning, and testing.
CURVE is NetEase's open‑source, high‑performance, highly available distributed storage platform that supports block, object, and cloud‑native database use cases, emphasizing simple solutions, hardware‑software integration, and extensive open‑source component reuse.
The Tencent Cloud CBS data scheduling system has evolved from a simple snapshot service into a highly concurrent, low‑latency platform that uses COW/ROW mechanisms, multi‑version snapshots, rapid rollback, hot‑data caching, horizontal scaling, fault‑tolerant task switching, cross‑region replication, and seamless disk migration to ensure reliable, fast storage for backups, image creation, and cloud‑disk migration, with future AI‑driven scheduling and ultra‑low‑latency features.
Tencent Cloud Block Storage (CBS) has evolved through three generations—apllo, atlas, and HiSTOR—adopting a client‑direct, distributed architecture, SPDK, RDMA and user‑space TCP to cut latency to sub‑microseconds while delivering exabyte‑scale throughput, high IOPS, and reliable multi‑copy replication for cloud VM workloads.
This article traces Ceph’s evolution from its academic origins to widespread adoption in cloud‑native environments, explains the four stages of storage development, details Ceph’s unified object, block, and file interfaces, and describes its architecture—including monitors, OSDs, CRUSH algorithm, placement groups, and high‑availability features.
NetEase Shufan has open‑sourced Curve, a high‑performance distributed storage system that claims 1.84× the IOPS of Ceph, offering low latency, high availability, and autonomous operation for cloud‑native workloads.
This article explains the fundamentals of Ceph, a high‑performance, highly available and scalable distributed storage system, covering its architecture, core components, data placement algorithms, storage interfaces, and typical deployment scenarios in cloud environments.
This article explains the core concepts of HDFS—including its block, packet, and chunk structures, their roles in data streaming, the detailed write and read workflows, and how checksums ensure data integrity—providing a comprehensive overview for anyone working with Hadoop distributed storage.
The July 2018 Ceph Developer Monthly report from the UMCloud storage team summarizes the latest community contributions, including enhancements to object and block storage, new OPA integration for fine‑grained access control, crash‑dump management, dashboard user UI, and batch operations for ceph‑volume.
The April 2018 Ceph monthly report highlights LTTng tracing for RGW, SSL support for the beast frontend, MFA integration, a notrim option for rbd mapping, runtime lz4 and brotli compression, Zabbix PG metrics, asynchronous dashboard tasks, detailed operation tracking, osdmap pruning, and a new iostat manager plugin.
This article presents an in‑depth overview of Zeppelin, a high‑availability, high‑performance block storage service, covering its motivation, online vs offline storage distinctions, data distribution strategies, centralized meta‑server design, replication policies, RocksDB‑based storage engine, Raft‑based consistency protocol, threading model, client request flow, and fault‑handling mechanisms.
This article details Tongcheng's adoption of Ceph for handling massive unstructured data, describing the motivations, architecture, specific uses of RGW, block storage, and CephFS, the challenges encountered, optimization measures, monitoring practices, and planned future improvements.
This article walks through the hardware and software preparation, configuration, deployment, testing, and troubleshooting steps required to set up a Ceph distributed block storage cluster on CentOS 7, including node roles, RAID setups, and Ceph‑deploy commands.
This article introduces Ceph as an open‑source distributed storage system that supports object, block, and file interfaces, explains its core RADOS components, outlines its advantages such as unified storage, high scalability, reliability and performance, and details how Ceph handles object and block storage within cloud environments.
FusionStorage is a distributed block storage solution for x86 servers that separates compute and storage, uses DHT‑based key‑value architecture, provides strong consistency and read‑repair mechanisms, and supports various cache media and advanced storage functions such as thin provisioning and synchronous replication.
