Inside Scality Ring: How Its Scale‑Out Architecture Powers Object, File, and Block Storage

Overview

Scality Ring is a software‑defined, scale‑out object storage system built on commodity X86 servers. Version 6.0 supports object, file, and block storage and integrates tightly with OpenStack, Hadoop, and various cloud platforms.

Three‑Layer Architecture

The system consists of three logical layers: a scalable object layer that implements efficient routing and replication, a distributed file/ block layer that exposes POSIX, NFS, SMB, FUSE, and Cinder interfaces, and a management layer providing GUI, CLI, and SNMP access.

Connectors

1. Scality Ring Connectors

Multiple connectors enable access to stored data:

Object Connectors : private high‑performance REST API, SNMP, S3, Swift, CDMI, and RS2/S3 for public‑cloud style access.

File Connectors : NFS v3 with quotas and advisory locks, SMB2.x/3.0 with encryption and failover, and FUSE (Sfused) for parallel I/O on large files.

Block Connectors : Cinder driver based on sparse files, plus a private block driver source.

Storage Nodes and I/O Daemons

Each storage server runs six Bizstorenode processes; each node hosts one Biziod daemon that communicates only with its Bizstorenodes. One Biziod maps to one physical disk, and a single server can host up to 255 Biziods. Containers are used to store metadata and object indexes, optionally backed by SSDs for fast metadata lookups.

System Management

Management is provided through a web‑based GUI, command‑line interface, and SNMP (MIB) support. The GUI offers dashboards showing zones, servers, and storage node status, while the CLI (RingSH) allows scripting of reliability policies and resource monitoring.

Scale‑Out File System (SOFS)

SOFS delivers a POSIX‑compatible file system without a separate gateway. It runs on top of a distributed NewSQL database called MESA, which maintains multiple copies of data for reliability and supports a volume abstraction similar to Gluster, allowing up to 2^32 volumes with billions of files each.

Routing Protocol and DHT

The ring architecture uses a chord‑style peer‑to‑peer algorithm. Lookup hops are bounded by ½·log₂(N), where N is the number of nodes. A custom 160‑bit key format supports up to six replicas or erasure‑coding (EC) groups, and updates are propagated only to affected nodes.

Intelligent Data Durability and Self‑Healing

Scality Ring offers multiple protection modes: simple replication (up to six copies, ideal for small files) and erasure coding (elastic EC for large files). The system can rebuild lost data from a single node failure within two hours, affecting only the failed node’s share of performance.

Multi‑Site Geo‑Distribution

Three deployment models enable geographic resilience:

Tier1Sync (S3 GeoBucket) : stretched ring across 2‑4 sites with <10 ms latency, providing synchronous or asynchronous writes.

Stretched SOFS : file‑system‑level multi‑site deployment where metadata updates occur locally and are asynchronously replicated.

Replicated SOFS : metadata is mirrored to independent rings at remote sites; the Ssync tool captures local metadata logs and replays them remotely, eliminating latency constraints.

Consistency Model

At the object layer, Ring follows an AP (available‑partition tolerant) model, offering eventual consistency. The SOFS layer adds strict POSIX semantics enforced by the MESA database, providing atomic file‑level operations.

Management Tools

Ring’s super‑management console is a web GUI that visualizes servers, zones, and storage nodes. It communicates with agents on management storage servers to collect statistics and perform health checks.

Solution Scenarios

Typical use cases include:

Active archive for long‑term data preservation and high‑value scientific datasets.

Content distribution with high bandwidth and unlimited scalability, competing with public‑cloud services.

Web and cloud services requiring unified object, file, and block interfaces.

Enterprise private clouds (SaaS storage platforms) that decouple software from hardware costs.

Scality Ring’s flexible architecture and extensive connector ecosystem make it suitable for a wide range of storage workloads across on‑premises and hybrid cloud environments.