OSS: The Giant Wheel Powering Cloud Object Storage

Object Storage Service (OSS) Overview

OSS stores unstructured data as objects in flat buckets without a directory hierarchy. Each object consists of data, metadata, and a globally unique object key (e.g., images/photo.jpg). A bucket can hold an unlimited number of objects; a single object can be up to 5 TB via multipart upload. Data durability is 99.999999999 % (11 9’s) through multi‑copy or erasure‑coding redundancy.

Key Features

Massive capacity – unlimited objects per bucket, EB‑level total storage.

Lifecycle management – rules based on time or prefix automatically transition objects to low‑frequency or archival storage, reducing cost up to 80 %.

Versioning – each write creates a unique version ID, enabling rollback and protection against accidental overwrite or deletion.

Cross‑region replication (CRR) – asynchronous incremental copy of bucket contents to another region for compliance and disaster recovery.

Static website hosting – HTML/CSS/JS files can be served directly from a bucket with custom domain binding, default index and error pages.

High availability – multi‑AZ redundancy provides 99.99 % service availability.

Pay‑as‑you‑go pricing – charges based on actual storage, request count, and outbound traffic.

S3‑compatible API – enables multi‑cloud migration and use of existing S3 tools.

Technical Principles

Data Model

Object = Data (binary content) + Metadata (type, size, creation time, etc.) + Object Key (unique identifier).

Distributed Storage Architecture

Data is split into chunks and stored across multiple physical nodes using a consistent‑hash algorithm. Redundancy is provided by default three‑copy replication or erasure‑coding (EC), which encodes fragments to recover from partial loss while improving storage efficiency.

Consistency Model

Upload and overwrite operations are strongly consistent – changes are immediately visible. Delete operations are eventually consistent, typically completing in less than one second.

Typical Application Scenarios

Static resource hosting – images, CSS/JS, fonts served from OSS and accelerated with CDN to offload origin servers.

Big‑data analytics – raw logs and sensor data stored in OSS can be processed directly by Hadoop or Spark; Alibaba Cloud OS + E‑MapReduce provides seamless HDFS integration.

Backup and archival – database backups and compliance archives stored in OSS; lifecycle policies move data to low‑frequency or archival tiers, saving up to 80 % of storage cost.

Multimedia processing – video‑on‑demand or real‑time image processing (scaling, watermarking) triggered by OSS events combined with Function Compute (e.g., Alibaba Cloud FC).

Access Methods

RESTful API – HTTP PUT/GET/DELETE; usable with curl, Postman, etc.

SDKs – Java, Python, Go and other language SDKs (Alibaba Cloud OSS SDK) encapsulate API calls.

Command‑line tools – ossutil, AWS CLI for bucket and object management.

Web console – graphical console for upload, download, and configuration.

Security Mechanisms

Identity authentication – AccessKey ID + Secret signs requests; STS issues temporary tokens with limited permissions and expiration.

Permission control – Bucket Policy (JSON) defines cross‑account or IP allow/deny lists; ACL provides coarse‑grained public‑read or private settings; RAM policies enable fine‑grained user/role permissions.

Data encryption – Server‑side encryption (SSE) with KMS‑managed keys or BYOK; client‑side encryption encrypts data before upload.

Logging and auditing – Access logs record all bucket operations and can be stored in a designated bucket; ActionTrail captures account‑level events for audit.

Architecture Diagrams