Scality Blog | Object storage, AI data infrastructure & cyber resilience

How should you store large datasets for AI training?

Written by Joshua Silvia | Sep 14, 2026, 3:42:07 PM

A training cluster is only as productive as the storage behind it. When accelerators sit idle because the data loader cannot read samples fast enough, or a checkpoint write stalls a job, the most expensive part of the infrastructure is waiting on the least visible part. Storing large datasets for AI training is therefore not a question of finding enough capacity, but of matching storage behavior to each stage of the AI data lifecycle.

Most organizations arrive at this problem gradually. A pilot runs against a departmental NAS share, the dataset grows to petabytes, and someone proposes copying everything into a public cloud where GPU capacity is available. Each step solves an immediate need and creates a structural one. This article walks through the lifecycle stages, why the usual starting points stop working, and how a two-tier architecture with object storage as the system of record addresses the whole lifecycle.

What is the AI data lifecycle?

The AI data lifecycle is the sequence of stages data passes through from collection to the point where a model built on it is retrained or retired. The naive view treats AI storage as a single problem: a fast disk pool next to the GPUs. In practice, data spends most of its life outside the training loop, and each stage asks something different.

  • Raw collection: ingesting images, video, sensor logs, documents or text from many sources. Volume, often petabytes, is the defining characteristic.
  • Curation and labeling: deduplicating, filtering and annotating raw data into training-ready datasets. Every sample acquires tags, provenance and version history.
  • Training: feeding curated datasets to accelerators across many epochs and nodes in parallel. Sustained read throughput is what counts.
  • Checkpoints: periodically writing model state so a job can resume after failure. Large, bursty, write-heavy.
  • Inference and fine-tuning: serving models, collecting new production data, and periodically retraining. Source data must remain traceable for as long as the model or its successors are in use.

What does each stage ask of storage?

Throughput to keep accelerators fed

Storage must deliver samples faster than accelerators consume them. Data loaders parallelize reads across many workers, so the pattern is many concurrent readers requesting small objects at random offsets across a very large dataset. Aggregate read bandwidth and concurrency without hot spots are the metrics that matter.

Capacity and long retention

Raw collections grow continuously and are rarely deleted, because the next model may need a different slice of the same source material, and scrutiny of AI systems increasingly asks what a model was trained on. Source data, labels and checkpoints may need to be retained for years with clear ownership and access control, so the capacity tier must scale to petabytes at a cost per terabyte that makes this viable.

Versioning and reproducibility

A model is reproducible only if the exact dataset it was trained on can be reconstructed later. That requires immutable snapshots, rich metadata, and the ability to reference a specific corpus version long after newer ones exist. Pipelines that overwrite files in place quietly destroy this property.

Checkpoint write bursts

Suppose a distributed job checkpoints every 30 minutes and each checkpoint is several hundred gigabytes. Every node writes its shard at the same moment, and the job cannot proceed until the write completes. Storage needs enough write bandwidth to absorb that burst quickly and enough durability that the checkpoint can be trusted weeks later.

Why do teams outgrow NAS?

Network-attached storage is where most AI projects begin: it is already there, it presents a familiar file interface, and it works well at small scale. A traditional NAS scales up rather than out, so adding capacity means adding shelves behind the same controllers, which become the bottleneck for throughput and metadata operations alike. Directory trees holding hundreds of millions of small files degrade lookup performance, and jobs contending for one controller pair see unpredictable latency that shows up as accelerator idle time. These are not defects; they are characteristics of an architecture built for a different workload.

Why does copying data to public cloud for every run create problems?

When on-premises storage cannot keep up, an obvious workaround is to copy the dataset into a public cloud region next to rented GPUs. This works once and becomes expensive and hard to govern as a habit.

The first issue is egress cost. Data moves in cheaply, but moving results, checkpoints or the dataset back out incurs per-gigabyte charges, and a team that repeats the cycle pays repeatedly. The second is governance: every copy is a new location where access must be controlled, retention enforced and, in regulated sectors, residency demonstrated. Partial copies in several regions make it hard to answer the auditor’s eventual question: where does this data live, who can reach it, and which version trained this model? The third is reproducibility: ad hoc copies drift, with no record of the difference.

How do the main storage architectures compare across the lifecycle?

Architecture Capacity tier fit Training throughput Metadata and versioning Cost model Data control
File / NAS Limited; scale-up controllers cap growth Adequate for small clusters; contention at scale Basic POSIX attributes; snapshots but weak dataset-level versioning Appliance capital cost, per-shelf expansion Full, on-premises
Parallel file system Possible but costly per terabyte at petabyte scale Very high; built for concurrent HPC-style reads and writes Strong file metadata; versioning left to external tooling High-performance media and specialist operation Full, on-premises
On-premises object storage Strong; scale-out to petabytes and exabytes High aggregate throughput for parallel readers; not the lowest single-request latency Rich object metadata, native versioning, immutability Software on commodity hardware, predictable per terabyte Full, on-premises or sovereign
Cloud object storage Effectively unlimited High within the same cloud; constrained across the network boundary Rich metadata, versioning, immutability Consumption pricing plus egress and API charges Shared responsibility; residency depends on provider region

No single row answers the whole lifecycle. Parallel file systems excel at training and are poorly suited to holding years of raw data. Object storage is the natural capacity and governance tier but not the lowest-latency path for a hot working set. Cloud object storage offers scale but raises cost and control questions once data crosses the boundary.

What is the hot tier plus capacity tier pattern?

Mature AI environments separate the two jobs the lifecycle demands. Object storage acts as the durable system of record for everything: raw collections, versioned datasets, checkpoints, model artifacts and retained production data. A fast tier, whether a parallel file system, local NVMe on the training nodes, or a cache in front of the object store, holds only the active working set for running jobs.

The two tiers have opposite economics. The capacity tier is optimized for cost per terabyte, durability and metadata richness; the hot tier for bandwidth and latency, kept small and transient. Data is staged from the capacity tier when a job starts, checkpoints are written back to it, and the hot tier is flushed when the job finishes. Several properties follow.

  • The system of record stays in one place under one access-control and retention policy, which simplifies audit and residency questions.
  • Dataset versions are immutable objects with metadata, so any training run can be tied to an exact corpus.
  • Training clusters, on-premises or rented, pull only the working set they need, which contains egress and duplication, and the fast tier can be replaced as accelerator generations change without migrating source data.

The S3 API is the practical glue: data loaders and checkpointing libraries increasingly speak S3 natively, so the object tier can be read directly where latency is less critical and staged to the hot tier only when needed.

What questions should a team ask before choosing?

  • How large is the raw corpus today, and how fast will it grow?
  • What aggregate read bandwidth keeps the cluster saturated, and how much of the corpus is active in any single job?
  • How large and frequent are checkpoints, and can every production model be traced back to an immutable, versioned dataset?
  • Where must the data reside, and if training moves to another site or provider, how much data moves with it and at what cost?

How Scality ADI and RING support AI training data

Scality ADI, the company’s autonomous data infrastructure platform, is organized around four pillars, two of which map directly to this lifecycle. MULTISCALE addresses the capacity tier: scale-out storage from multi-petabyte to exabyte scale, so raw collections and long-retention source data stay in one system of record. AICONNECT addresses the link between that system of record and the training environment, so pipelines and accelerators reach versioned datasets through standard S3 and file interfaces rather than ad hoc copies.

Underneath, Scality RING provides the software-defined object and file storage that serves as the durable tier: an S3 API alongside file access, object versioning and immutability for reproducibility and audit, on commodity hardware on-premises or in a sovereign environment where the organization retains full control of its data. Scality describes the combination on its AI core use case page and defines the terms in its AI storage glossary entry.

The practical takeaway: decide first where the single, versioned system of record for AI data will live, and treat the fast tier as a replaceable accessory to it.