AWS S3, Microsoft Azure Blob Storage and Backblaze B2 are object-storage services that can support media archives, backups and application workflows. Their relative performance and cost depend on region, storage class, object size, request pattern, concurrency, retrieval frequency, egress, redundancy and contract terms. This article provides a reproducible evaluation framework. It does not report a completed benchmark or rank the providers.
Define the storage workload
Describe the intended use: camera-original archive, proxy delivery, active editing, render output, backup or application asset store. Record the number of objects, size distribution, monthly growth, GET/PUT/LIST/HEAD request volume, retrieval frequency, concurrency and required recovery objectives. These workloads have different performance and cost profiles.
For the storage architecture and recovery-cost considerations behind NAS and cloud DAM, see NAS versus cloud DAM.
Compare equivalent service configurations
For each provider, record the exact region, storage class or tier, redundancy option, lifecycle policy, encryption configuration, API endpoint, account type and date of pricing review. Include minimum storage duration, retrieval fees, early deletion charges, request charges, egress and any transfer acceleration or CDN services used. Do not compare a low-cost archive tier with a hot tier as if they provide the same service.
For independent backup copies, retention and restore testing, see our multi-terabyte backup architecture guide.
Reproducible benchmark protocol
- Use the same client hardware, network path and software runtime for each provider.
- Test multiple object sizes, from small metadata files to large media masters.
- Measure upload and download throughput at several concurrency levels.
- Record P50, P95 and P99 request latency, errors, retries and timeouts.
- Separate cold and warm tests and document cache conditions.
- Test multipart uploads, range reads, listing and version retrieval where relevant.
- Repeat tests in each region that the organisation may actually use.
Build a total-cost model
Calculate the expected monthly cost as storage capacity × storage rate, plus request charges, retrieval charges, egress, replication, backup, support and operational tooling. Use the provider’s dated pricing calculator or quote for the selected region and storage tier. Include data growth and expected retention across the full evaluation period.
For the local storage bottlenecks that can affect media workloads, see NVMe storage in AI editing workflows.
Security and resilience comparison
Compare identity and access controls, encryption options, versioning, object retention, immutability, audit logging, replication and restore behaviour. Confirm whether the selected storage tier supports the required features. For example, Amazon S3 Object Lock requires versioning and protects object versions under configured retention or legal-hold rules; it is not a substitute for testing restoration. See AWS S3 Object Lock documentation.
Decision criteria
Choose the service that best meets the required workload, recovery objectives, regional availability, security controls and total cost. A provider that is least expensive for a cold archive may not be the best choice for frequent random access or high-volume egress. Keep benchmark results tied to the tested region, account, date and configuration.
Conclusion
AWS S3, Azure Blob and Backblaze B2 should be compared using equivalent storage tiers, regions and request patterns. Until controlled measurements and dated cost calculations are published, no performance or cost winner is claimed. The framework above can be used to build a defensible provider decision from the organisation’s own workload.