What Is Amazon S3 Glacier and How to Use It

Summary:

• Amazon S3 Glacier encompasses three storage classes (Instant Retrieval, Flexible Retrieval, and Deep Archive) designed for long-term archival data with varying retrieval speeds and cost structures.
• The standalone Amazon Glacier vault service is a legacy architectural pattern, with AWS strongly recommending all new and existing workloads migrate to the unified Amazon S3 Glacier storage classes via the S3 API.
• Lifecycle policies automate data transitions between storage tiers, enabling organizations to optimize costs while maintaining compliance with retention requirements.
• Deep Archive offers the lowest storage cost at approximately $0.00099 per GB-month but requires up to 12 hours for standard retrieval, making it ideal for data accessed once or twice per year.

Storing petabytes of compliance records, medical imaging archives, or media production assets demands a fundamentally different cost calculus than serving hot application data. Amazon S3 Glacier storage classes address this reality by offering archive storage engineered for durability measured in eleven nines while keeping monthly bills a fraction of standard S3 costs.

The landscape has shifted dramatically. The original standalone Amazon Glacier vault service is now considered a legacy pattern. Understanding how to navigate the unified S3 Glacier ecosystem separates cost-optimized architectures from expensive technical debt. This guide unpacks each storage class, compares retrieval economics, walks through migration from legacy vaults, and provides the lifecycle policy configurations that enterprise teams need to implement immediately.

What is Amazon S3 Glacier and the status of standalone vault service

Amazon S3 Glacier represents a family of storage classes within Amazon S3 purpose-built for data archiving and long-term backup. Unlike standard S3 storage optimized for frequent access, S3 Glacier storage classes trade retrieval speed for dramatically lower storage costs, achieving prices as low as $0.00099 per GB-month in US-East regions.

All three Glacier classes deliver 99.999999999% (eleven nines) durability by automatically replicating data across a minimum of three Availability Zones within a region. Objects stored in Glacier classes remain fully managed through the familiar S3 API. They support features like object tagging, metadata (up to 2 KB of user-defined metadata per object), and integration with S3 Object Lock for WORM (Write Once Read Many) compliance requirements.

The distinction between Amazon S3 Glacier and the original Amazon Glacier service causes significant confusion. The standalone Amazon Glacier vault service operated as a separate AWS service with its own API, vault-based organization, and archive-centric data model. While AWS announced that this standalone vault-based service will no longer accept new customers starting December 15, 2025, existing customer data remains secure and accessible indefinitely. There is no forced migration or end-of-life deadline for existing vaults.

The consolidation under S3 brings operational advantages that justify the migration effort. Teams now manage archival data using the same IAM policies, bucket configurations, and monitoring tools as their primary S3 workloads. S3 Inventory reports, S3 Storage Lens analytics, and CloudWatch metrics provide unified visibility across all storage classes.

This architectural simplification reduces operational overhead while enabling sophisticated lifecycle automation that the standalone vault service never supported. The following section provides a detailed breakdown of each storage class and their specific characteristics.

S3 Glacier storage classes overview

Amazon S3 Glacier encompasses three distinct storage classes, each optimized for different access patterns and cost sensitivities. Selecting the appropriate class requires understanding the interplay between storage costs, retrieval costs, retrieval latency, and minimum storage duration commitments. Objects smaller than 128 KB incur minimum billable size charges, making Glacier classes most cost-effective for larger objects or batched small files. The following comparison table provides the essential metrics for architectural decision-making.

Glacier Instant Retrieval

Glacier Instant Retrieval targets data accessed approximately once per quarter but requiring immediate availability when needed. Medical imaging archives exemplify this pattern. A hospital might access historical radiology scans infrequently, yet clinicians require millisecond retrieval when reviewing patient history during diagnosis.

Storage costs run approximately 68% lower than S3 Standard-Infrequent Access while maintaining the same millisecond first-byte latency. The 90-day minimum storage duration means objects deleted or transitioned before 90 days incur pro-rated charges for the remaining duration.

Glacier Flexible Retrieval

Glacier Flexible Retrieval (formerly known simply as “Amazon S3 Glacier”) provides the middle ground for data accessed once or twice per year where retrieval can tolerate minutes to hours of latency. This class offers three retrieval tiers. Expedited takes 1-5 minutes at the highest cost. Standard takes 3-5 hours at moderate cost. Bulk takes 5-12 hours at the lowest cost.

Organizations archiving financial records for regulatory compliance often choose Flexible Retrieval because audits are infrequent and can accommodate multi-hour retrieval windows. The 90-day minimum storage duration matches Instant Retrieval, but storage costs drop to $0.0036 per GB-month.

Glacier Deep Archive

Glacier Deep Archive delivers the lowest storage cost in the AWS portfolio at $0.00099 per GB-month. It is designed for data retained seven to ten years or longer with extremely rare access requirements. Retrieval takes 12 hours for Standard tier or up to 48 hours for Bulk operations.

The 180-day minimum storage duration reflects the class’s positioning for truly cold data. Regulatory archives, genomic research datasets, and historical surveillance footage represent typical Deep Archive workloads where the cost savings justify extended retrieval windows.

Understanding these storage class characteristics enables informed lifecycle policy design. The next section examines specific use cases with cost projections that demonstrate the financial impact of storage class selection over multi-year retention periods.

When to use each storage class

Selecting the optimal S3 Glacier storage class requires mapping your data access patterns against the cost and retrieval characteristics of each tier. The decision extends beyond simple storage cost comparison to encompass retrieval frequency, latency tolerance, and total cost of ownership across your retention period. Consider a healthcare organization archiving 500 TB of medical imaging data with a seven-year retention requirement and quarterly access patterns for approximately 1% of the archive.

Cost analysis for medical imaging archive

Using Glacier Instant Retrieval for this workload produces the following five-year cost projection:

• Annual storage cost: 500 TB × $0.004/GB × 12 months = $24,000
• Annual retrieval cost: 5 TB retrieved × 4 quarters × $0.03/GB = $600
• Five-year total: Approximately $123,000

Switching to Glacier Flexible Retrieval with Standard tier retrievals changes the economics significantly. Storage drops to $21,600 annually, but the organization must accept 3-5 hour retrieval delays. For clinical workflows requiring immediate image access, this latency proves unacceptable despite the cost savings. Research departments accessing the same archive for retrospective studies might find Flexible Retrieval perfectly suitable.

Compliance and regulatory archives

Financial services organizations face strict data retention mandates, often requiring seven-year preservation of transaction records and communications. Deep Archive excels here because regulatory audits are rare events with predictable timelines. A bank archiving 100 TB of compliance data realizes the following annual costs:

• Deep Archive storage: 100 TB × $0.00099/GB × 12 months = $1,188
• Flexible Retrieval storage: 100 TB × $0.0036/GB × 12 months = $4,320
• Seven-year savings with Deep Archive: Approximately $21,924

The 180-day minimum storage duration for Deep Archive rarely impacts compliance workloads since retention periods span years. Organizations must factor retrieval costs into audit response planning. Restoring 10 TB for an audit using Deep Archive Standard retrieval adds $200 in retrieval fees plus the operational overhead of 12-hour wait times.

Data lake cold tier optimization

Modern data lakes benefit from tiered storage strategies that automatically transition aging data to cost-appropriate classes. Raw ingestion data might reside in S3 Standard for 30 days during active analysis, transition to S3 Standard-IA for 60 days, move to Glacier Instant Retrieval for 180 days, and finally settle in Deep Archive for long-term retention. This progression optimizes costs while maintaining accessibility aligned with actual usage patterns. The lifecycle policy configuration enabling this automation forms the focus of the following section.

Migrating from original Glacier vaults and lifecycle policy setup

While organizations operating legacy Amazon Glacier vaults are not subject to a mandatory migration, moving to S3 Glacier storage classes is highly recommended to unlock modern unified management capabilities. The migration process involves creating destination S3 buckets, initiating vault inventory retrieval, downloading archives, and uploading to S3 using the appropriate Glacier storage class. AWS provides the Glacier archive retrieval documentation detailing the two-step process of initiating jobs and downloading output.

Migration roadmap for vault users

The migration timeline requires careful planning given the retrieval latencies inherent to Glacier vaults. Follow this sequence for successful migration:

1. Inventory your vaults: Initiate vault inventory jobs to catalog all archives. Inventory retrieval takes 3-5 hours for vaults with fewer than 10,000 archives.
2. Calculate transfer time: Estimate total archive size and plan retrieval batches. Bulk retrievals minimize costs but extend timelines to 5-12 hours per batch.
3. Create destination buckets: Configure S3 buckets with appropriate storage class defaults and enable versioning if required for compliance.
4. Execute staged migration: Retrieve archives in batches, upload to S3 using multipart upload for objects exceeding 100 MB, and verify integrity using checksums.
5. Validate and decommission: Confirm all archives exist in S3, update application configurations to use S3 APIs, and delete empty vaults.

Configuring S3 lifecycle policies

Lifecycle policies automate object transitions between storage classes based on age, eliminating manual data management overhead. Policies attach to S3 buckets and evaluate objects against defined rules daily. The following JSON configuration demonstrates a multi-tier lifecycle policy transitioning objects through progressively colder storage classes.

Key configuration considerations include:

• Prefix and tag filters: Apply transitions selectively using object key prefixes or tags rather than bucket-wide rules.
• Minimum object size: S3 Lifecycle rules now apply a default 128 KB minimum size filter for transitions (which can be overridden). However, objects moved to Glacier Flexible Retrieval or Deep Archive incur a 40 KB metadata overhead per object. Always batch small files before transitioning, as millions of small files will balloon your storage costs through this metadata tax.
• Transition timing: The 30-day minimum storage requirement applies only when transitioning from S3 Standard to Standard-IA or One Zone-IA. You can configure lifecycle policies to transition objects directly from S3 Standard to any Glacier class immediately (Day 0) if your architecture demands it.

The AWS lifecycle transition documentation provides comprehensive guidance on transition constraints and cost implications. With lifecycle automation established, securing archived data becomes the next critical consideration.

Security and compliance for S3 Glacier

S3 Glacier storage classes inherit the comprehensive security framework of Amazon S3, including encryption, access control, and audit capabilities. All data stored in Glacier classes receives server-side encryption by default using S3-managed keys (SSE-S3).

For advanced encryption, AWS KMS-managed keys (SSE-KMS) are the recommended standard. While customer-provided keys (SSE-C) were historically available, as of April 2026, AWS disables SSE-C by default on all new S3 buckets. AWS implemented this change because SSE-C lacks the flexibility required for modern analytics and ML integrations; teams requiring SSE-C must now explicitly unblock it via the PutBucketEncryption API. Encryption applies transparently during the S3 PUT API operation and decryption occurs automatically during retrieval.

Access control operates through IAM policies, bucket policies, and S3 Access Points. Organizations implementing least-privilege access should restrict Glacier restore operations to specific IAM roles, preventing unauthorized data retrieval that could incur unexpected costs. The following policy elements control Glacier-specific operations:

• s3:RestoreObject: Permission required to initiate retrieval from any Glacier storage class
• s3:GetObject: Permission required to download restored objects during the restoration window
• s3:PutObjectStorageClass: Permission required to transition objects between storage classes

Compliance certifications

Amazon S3 Glacier storage classes maintain compliance with major regulatory frameworks and industry standards. AWS publishes compliance reports through AWS Artifact, providing auditor-ready documentation for:

• SOC 1, SOC 2, SOC 3: Service organization control reports covering security, availability, and confidentiality
• HIPAA: Healthcare data protection when used with a signed Business Associate Agreement
• PCI DSS: Payment card industry compliance for cardholder data archives
• FedRAMP: Federal government authorization for AWS GovCloud regions

S3 Glacier Deep Archive specifically addresses long-term retention mandates in financial services (SEC Rule 17a-4), healthcare (HIPAA), and government (NARA) sectors. The combination of Object Lock, CloudTrail logging, and S3 Access Logs provides the audit trail necessary for demonstrating compliance during regulatory examinations.

Conclusion

Amazon S3 Glacier storage classes provide the foundation for cost-optimized archival storage, with each tier addressing distinct access patterns and latency tolerances. Glacier Instant Retrieval serves quarterly-access workloads requiring millisecond response. Flexible Retrieval handles annual-access patterns with hour-scale retrieval. Deep Archive delivers sub-penny-per-gigabyte storage for decade-scale retention. While AWS stops accepting new customers for standalone Glacier vaults in late 2025, migration to S3 storage classes is highly recommended for organizations looking to modernize their legacy archives and unify management.

Lifecycle policies transform manual storage management into automated cost optimization, transitioning objects through storage tiers as access patterns evolve. Combined with S3 Object Lock for compliance immutability and comprehensive IAM controls, Glacier storage classes satisfy both economic and regulatory requirements. Organizations architecting data lakes, compliance archives, or media asset management systems should evaluate their access patterns against the retrieval characteristics and minimum storage durations of each class. This approach minimizes total cost of ownership while maintaining operational flexibility.