Glossary

Software-defined infrastructure

Software-defined infrastructure (SDI) is a data centre model in which compute, storage and networking are separated from their physical hardware and provisioned, configured and monitored through software interfaces. Resources are pooled across standard servers and switches and allocated by API, usually under automation.

Why software-defined infrastructure matters for platform teams

Platform teams in large organizations, and service providers selling capacity, deliver infrastructure as an internal or external service. Requests arrive in the hundreds per week from pipelines, developers and tenants. When each request means a ticket for a server, another for a storage volume and a third for network rules, the slowest team sets the pace for everyone. SDI turns each of those steps into an API call that automation can issue in seconds.

Storage is frequently the last layer to make that transition, because it holds state that outlives any workload. An estate where compute and networking are fully automated but storage still needs manual provisioning keeps a manual bottleneck at its centre.

Compute, storage and network as software

Each layer has a data plane that carries the work and a control plane that decides how it is configured. SDI exposes the control planes through programmable interfaces, typically REST APIs with resources described in JSON or YAML.

Declarative configuration and reconciliation

SDI systems commonly accept a description of the desired state rather than a sequence of commands: four instances, each with 2 vCPUs, 8 GB of memory and a 100 GB volume, on a named network. A controller compares the declaration with what is running and corrects any difference, so if one instance fails it starts a replacement. Kept under version control, these declarations are infrastructure as code, and the same files can rebuild an environment at another site, which links SDI to cloud automation.

How storage is consumed in SDI

• Block volumes are requested by the scheduler through a storage driver, attached to a VM or container when it starts and reattached elsewhere if it moves.
• Shared file systems are mounted over NFS or SMB by workloads that need a common directory tree.
• Object storage is reached over S3 with no attachment step: any workload holding credentials and network access can read and write, which makes it the usual home for state shared across many short-lived instances.

A block volume follows its workload; a bucket persists independently of every instance. Stateless application designs keep instance disks disposable and hold durable state in databases and object stores.

Object storage access is governed by credentials and policies, so tenancy in SDI is expressed as accounts, buckets, access keys and bucket policies issued by automation alongside the compute that uses them. When a project is torn down, the same automation can revoke its keys while the data it produced stays in place for whoever needs it next.

What software-defined infrastructure means for the storage layer

For storage teams inside an SDI estate, the measure of the platform changes. Provisioning has to happen by API at the speed compute is scheduled, with quotas, protection classes and access policies applied automatically per tenant or project. Capacity becomes a pooled, metered resource, and the storage team's work shifts from carving volumes to setting the policies automation applies.

Failure domains become logical. A placement policy that spreads copies across racks or sites is enforced by software, and a misconfigured policy can place every copy of a data set behind the same top-of-rack switch without any visible change in day-to-day behaviour. Upgrades roll across nodes while workloads keep running, which depends on each layer tolerating the temporary loss of a node. Short-lived GPU jobs and CI pipelines that start, read terabytes and stop again turn storage into the one persistent component, so its throughput under many parallel clients sets the pace for the rest of the platform. Object storage tends to grow fastest in these environments because cloud-native and AI workloads write their durable state to S3. SDI is the technical base for private cloud and infrastructure as a service, which add self-service and billing on top, and hyper-converged infrastructure is one way to arrange the hardware beneath it.

Scality storage in software-defined infrastructure

Scality RING provides the storage layer as software on standard x86 servers, serving S3, NFS and SMB as listed on its MultiScale architecture page, with protection set per storage class and multi-tenancy with hard isolation. Scality's Autonomous Data Infrastructure (ADI) builds on the same distributed object storage with S3 and S3 over RDMA data paths, deployable on premises, in a sovereign cloud or as a managed service.