Meter
A retrieval tier is a speed option, not a storage class
A retrieval tier is a speed option for making archive data readable. It sits inside a storage class's recovery workflow and can change latency, retrieval price, and restore-request price without changing where the source object is stored.
Put your own numbers through the calculator to see what this meter does to a normal month and to the day you restore.
A retrieval tier chooses recovery speed inside an archive workflow
A retrieval tier is a named speed option for making archive data readable. The selected option can change the expected delay, price per retrieved byte, and price per restore request. [3] [4] [5]
A retrieval tier is not a storage class. The object remains in its selected class, while the retrieval tier describes how one recovery request is processed. A provider that publishes one retrieval rate with no speed choice has a retrieval meter but no named retrieval-tier choice in the captured data. [3] [1]
| Term | What it selects | What can change | Source ref |
|---|---|---|---|
| Storage class | The object's ongoing storage and access contract [2] | Storage, retrieval, requests, duration, overhead, and restore workflow | [2] [4] |
| Retrieval tier | The speed mode for one archive recovery request [3] | Latency, retrieval bytes, and restore-request price | [3] [4] |
| Retrieval latency | The published time range before data becomes readable [3] | Recovery-time eligibility | [3] |
| Egress | The later network path to the recovery destination [4] | Destination-aware transfer cost, separate from retrieval | [4] |
Amazon S3 and Azure publish named speed choices in the captured dataset
Only Amazon S3 and Azure Blob Storage publish multiple named retrieval-speed options in the 61 captured offers. The table copies the rates, requests, regions, and latency wording from providers.json and its official source pages; it does not recompute them. [1] [6] [7] [8]
A numeric zero copied from a published tier is rendered as not charged. Missing latency remains not published, and one named option is not presented as a multi-option choice. [1]
Amazon S3 and Azure publish named speed choices in the captured dataset9 rows
| Provider and class | Retrieval tier | Retrieved-byte treatment | Restore-request treatment | Published latency | Source ref |
|---|---|---|---|---|---|
| Amazon S3 Glacier Flexible Retrieval, us-east-1 | Expedited | $0.03/GB [6] | $10.00 per 1,000 [6] | 1 to 5 minutes [3] | [6] [3] |
| Amazon S3 Glacier Flexible Retrieval, us-east-1 | Standard | $0.01/GB [6] | $0.05 per 1,000 [6] | 3 to 5 hours [3] | [6] [3] |
| Amazon S3 Glacier Flexible Retrieval, us-east-1 | Bulk | not charged [6] | not charged [6] | 5 to 12 hours [3] | [6] [3] |
| Amazon S3 Glacier Deep Archive, us-east-1 | Standard | $0.02/GB [6] | $0.10 per 1,000 [6] | Within 12 hours [3] | [6] [3] |
| Amazon S3 Glacier Deep Archive, us-east-1 | Bulk | $0.0025/GB [6] | $0.025 per 1,000 [6] | Within 48 hours [3] | [6] [3] |
| Azure Blob Archive LRS, East US | Standard rehydration | $0.02/GB [7] | $0.50 per 1,000 [7] | Up to 15 hours for objects under 10 GB [5] | [7] [5] |
| Azure Blob Archive LRS, East US | High-priority rehydration | $0.10/GB [7] | $5.00 per 1,000 [7] | May complete in less than 1 hour for objects under 10 GB [5] | [7] [5] |
| Azure Blob Archive LRS, West Europe | Standard rehydration | $0.024/GB [8] | $0.60 per 1,000 [8] | Up to 15 hours for objects under 10 GB [5] | [8] [5] |
| Azure Blob Archive LRS, West Europe | High-priority rehydration | $0.13/GB [8] | $6.50 per 1,000 [8] | May complete in less than 1 hour for objects under 10 GB [5] | [8] [5] |
Most vendors publish a single retrieval rule or no named speed tier
The groups below are class-scoped and can overlap for a provider with several classes. They preserve the difference between a named speed choice, one paid retrieval rule, an allowance with an unpublished overage, a not charged rule, and missing archive details. [1]
A read-only restore can add retrieval, restore and read requests, temporary-copy storage, and egress. It never triggers early deletion. [16] [4]
| Captured retrieval shape | Providers or classes | State treatment | Source ref |
|---|---|---|---|
| Multiple named speed choices | Amazon S3 Glacier Flexible Retrieval and Deep Archive; Azure Blob Archive [1] | Use the selected tier's byte, request, and latency fields | [1] |
| Single paid retrieval rule with no named speed choice | Google Nearline, Coldline, and Archive; R2 Infrequent Access; Scaleway Glacier; OVHcloud Infrequent Access and Active Archive; Oracle Infrequent Access; Tigris Infrequent Access and Archive Instant Retrieval [1] | A retrieval meter exists, but retrieval-tier latency is not published unless separately sourced | [1] |
| Allowance with priced excess | DigitalOcean Spaces Cold Storage [9] | $0.01/GiB after the monthly 1x average-daily-storage allowance; recovery latency remains not published | [1] [9] [10] |
| Archive retrieval details incomplete | OVHcloud Cold Archive; Vultr Archive [11] [12] | not published; do not name a cheapest or fastest option | [11] [12] |
| No paid retrieval tier in the captured primary offer | Backblaze B2; Wasabi; IDrive e2; Hetzner; Akamai; captured Storj Standard; Filebase; Telnyx [1] | not charged for the captured retrieval meter where stated, not a claim about every future class | [1] |
| Cloudflare R2 recovery boundary | Infrequent Access retrieval is $0.01/GB while internet egress is not charged; Class A and Class B operations remain billed [13] | A zero egress line does not erase retrieval or requests | [13] |
| Wasabi recovery boundary | Retrieval is not charged in the captured offer, while excess egress pricing is not published beyond the fair-use policy [1] [14] | Stop at not published rather than turn excess into $0 | [1] [14] |
| Backblaze B2 recovery boundary | Retrieval is not charged; egress is free up to 3 times average monthly storage, then $0.01/GB [1] [15] | Apply the published allowance before the overage | [1] [15] |
Sources
Every figure above comes from one of these pages, on the date shown. Each numbered reference in the text links to its entry here. Vendors change prices; if a date looks old, check the source. Three vendors publish a machine-readable feed and the rest are re-checked by hand, which is why the dates are not uniform.
The source ledger16 sources
- 1This repository: providers.jsonrepository · checked 2026-07-25T15:34:44Z
- 2AWS documentation: Storage class introdocs.aws.amazon.com · checked 2026-07-26
- 3AWS documentation: Restoring objects retrieval optionsdocs.aws.amazon.com · checked 2026-07-26
- 4Amazon Web Services: S3 pricingaws.amazon.com · checked 2026-07-26
- 5Microsoft Learn: Archive rehydrate overviewlearn.microsoft.com · checked 2026-07-26
- 6AWS Price List API: Us east 1pricing.us-east-1.amazonaws.com · checked 2026-07-25T15:34:44Z
- 7Azure Retail Prices API: Retail pricesprices.azure.com · checked 2026-07-25T15:34:44Z
- 8Azure Retail Prices API: Retail pricesprices.azure.com · checked 2026-07-25T15:34:44Z
- 9DigitalOcean: Spaces object storagewww.digitalocean.com · checked 2026-07-26
- 10DigitalOcean documentation: Pricingdocs.digitalocean.com · checked 2026-07-26
- 11OVHcloud: Cold archivewww.ovhcloud.com · checked 2026-07-25T15:34:44Z
- 12Vultr documentation: Limitsdocs.vultr.com · checked 2026-07-26
- 13Cloudflare developer docs: R2 pricingdevelopers.cloudflare.com · checked 2026-07-25T15:34:44Z
- 14Wasabi: Faqwasabi.com · checked 2026-07-25T15:34:44Z
- 15Backblaze: Transaction pricingwww.backblaze.com · checked 2026-07-25T15:34:44Z
- 16AWS documentation: Restoring objectsdocs.aws.amazon.com · checked 2026-07-26