What Is S3 Glacier in Databases?

Imagine you have a big box of old tax documents from the past ten years. You don’t need to look at them every day, but you have to keep them by law just in case the government asks for them. Putting them in a regular office filing cabinet wastes expensive space that you could use for current projects.

Instead, you rent a small storage unit in a climate-controlled warehouse across town. That unit costs much less per month than your office space. But if you need a specific document, you have to drive to the warehouse, find the box, and bring it back.

The whole trip takes a few hours. That is exactly how Amazon S3 Glacier works. It is a place in the cloud where you can keep huge amounts of data for a very low monthly fee. The catch is that when you want that data back, you have to wait, sometimes minutes, sometimes hours, sometimes even overnight, before you can download it.

This trade-off makes sense for archived data. Think about medical records, scientific research data, or old surveillance footage. You almost never need them, but if you lose them, the consequences could be huge.

S3 Glacier is built for that exact scenario. It offers three retrieval options: Expedited (1–5 minutes), Standard (3–5 hours), and Bulk (5–12 hours). The faster you need the data, the more you pay per retrieval.

But the storage cost remains remarkably low, often a fraction of a cent per gigabyte per month. This makes S3 Glacier one of the most cost-effective solutions for long-term data retention in the cloud. For businesses that must comply with regulations like GDPR or HIPAA, S3 Glacier also provides encryption at rest and in transit, along with features like S3 Object Lock to prevent deletion or modification of files.

So while it is not suitable for active data, it is the perfect home for your digital attic.

Amazon S3 Glacier is a storage class within the Amazon S3 ecosystem, designed specifically for data archiving with retrieval times ranging from minutes to hours. It is part of the S3 Intelligent-Tiering family but is a distinct storage class optimized for long-term, infrequent access. S3 Glacier stores data as objects within S3 buckets, and you can transition objects into Glacier either directly by specifying the storage class at upload, or through lifecycle policies that automatically move older data from S3 Standard or S3 Standard-IA into Glacier after a defined period.

Under the hood, S3 Glacier uses a distributed, redundant storage architecture. Data is automatically replicated across at least three Availability Zones within the same AWS Region, ensuring 99.999999999% durability (eleven 9s). This means that even if an entire data center fails, your data remains safe. The service also supports server-side encryption with either S3-managed keys (SSE-S3), AWS Key Management Service keys (SSE-KMS), or customer-provided keys (SSE-C). You can enable S3 Object Lock to enforce a write-once-read-many (WORM) model, which is critical for compliance with regulations like SEC Rule 17a-4.

Retrieval is not instant. Glacier offers three retrieval tiers: Expedited (1–5 minutes, higher cost), Standard (3–5 hours, moderate cost), and Bulk (5–12 hours, lowest cost). Each tier has a different pricing model based on the speed of access. You can initiate a retrieval request through the AWS Management Console, AWS CLI, or SDK. Once the request is submitted, AWS prepares a copy of the data in a temporary staging area, and you can download it from there. You are not charged for data transfer into Glacier, but you are charged for data transfer out to the internet, as well as for retrieval requests and early deletion fees if you delete an object within 90 days of uploading it.

S3 Glacier also supports vaults and archives for the Glacier API (as distinct from the S3 API). In the Glacier API, a vault is a container for archives, similar to a bucket for objects. However, in practice, most users interact with Glacier through the S3 API, where Glacier is simply another storage class. AWS recommends using the S3 API for new projects because it provides a unified experience and access to advanced features like lifecycle policies, replication, and event notifications.

One important technical detail is the concept of "restore." When you request data from Glacier, it is not immediately available. Instead, you initiate a restore operation that copies the data to a temporary S3 storage class, such as S3 Standard. After the restoration completes, you can access the object via the S3 endpoint. You can specify how long the temporary copy should be kept (e.g., 30 days), after which it is automatically deleted. This process decouples the permanent archival storage from the temporary retrieval layer, keeping costs low while still allowing access when needed.

For IT professionals, implementing Glacier involves setting up lifecycle rules in S3. For example, you might configure a rule that moves objects older than 365 days from S3 Standard to S3 Glacier. You can also set up a transition to S3 Glacier Deep Archive for data that must be kept for 5+ years but accessed extremely rarely. Automation through AWS Lambda and EventBridge can notify administrators when restores complete or when data transitions between storage classes.

Security is built in at every layer. Data in transit is protected by TLS/SSL. Data at rest is encrypted by default. Access is controlled through IAM policies, bucket policies, and S3 Access Points. You can also enable S3 Access Logs to audit who accessed what and when. Glacier also supports VPC endpoints for private network access, and AWS Config can track configuration changes to your Glacier vaults and retrieval policies.

In terms of performance, the latency of retrieval depends on the tier and the size of the archive. Large archives take longer to restore because they need to be reassembled from multiple chunks. AWS uses a queuing system to handle retrieval requests, so during peak demand you might experience slightly longer wait times. For Expedited retrievals, you can also provision a certain amount of capacity (in MB/s) to guarantee faster access, which is useful for disaster recovery scenarios where you need a specific file quickly.

Finally, Glacier integrates with AWS Backup, Amazon S3 Lifecycle Management, and AWS Storage Gateway, making it a versatile option for hybrid cloud architectures. On-premises backup software, like Veeam or NetBackup, can also write directly to Glacier via the S3 API, enabling offsite vault copies for disaster recovery.

In the real world of IT, data grows exponentially every year. Companies generate terabytes of logs, emails, backups, and media files. Keeping all of that on expensive SSDs or even traditional hard drives would cost a fortune.

S3 Glacier matters because it gives organizations a legally compliant, highly durable, and drastically cheaper alternative for data that does not need immediate access. For example, healthcare providers must retain patient records for 7–10 years after the last visit. Storing those records on active storage would cost tens of thousands of dollars monthly.

Moving them to Glacier after a few months of activity can reduce storage costs by 80% or more. Financial institutions similarly need to keep transaction records for regulatory audits. Many use Glacier with Object Lock to automatically prevent deletion during the retention period.

E-commerce companies also use Glacier to archive old customer orders and invoices. While customers rarely request an invoice from three years ago, the company must keep the data for tax purposes. Glacier makes that economically viable.

Glacier is elastic. You do not need to forecast how much archival space you will need in ten years. You simply upload data as it ages, and AWS handles the capacity. The practical implementation often involves lifecycle policies.

For instance, a company might keep logs on S3 Standard for 30 days, move them to Standard-IA for 90 days, then transition to Glacier for 7 years. At the end of the 7-year period, the logs automatically expire and are deleted. This entire process requires minimal human intervention, saving time and reducing errors.

Glacier also enables hybrid cloud backup. Using AWS Storage Gateway, you can point your on-premises backup software to an S3 bucket that transitions data to Glacier. If your office burns down, your archives are safe in the cloud and accessible from anywhere.

Without Glacier, many small and medium businesses would be forced to keep expensive tape backup systems or risk losing critical data. So, S3 Glacier is not just a storage service; it is a fundamental tool for cost management, compliance, and data lifecycle automation in modern IT.

In AWS certification exams, questions about S3 Glacier typically fall into three categories: scenario-based, configuration-based, and comparison-based. In scenario-based questions, you are given a description of a company’s data usage pattern and need to choose the right storage class. For example: A financial firm must retain three years of transaction logs.

The logs are accessed only once a year for auditing. Which storage class provides the lowest cost while meeting retrieval needs? The answer is S3 Glacier, because retrieval within hours is acceptable and cost is the priority.

Another scenario: A media company needs to archive raw video footage for five years. Occasionally, an editor needs a specific clip within 5 minutes to respond to a court order. Which retrieval tier should they choose?

Expedited retrieval, because speed is critical. Configuration-based questions involve lifecycle policies. You might see a JSON snippet of a lifecycle rule and you have to identify what it does.

For instance, a rule that transitions objects to Glacier after 30 days and expires them after 365 days. The question might ask: How long does the object stay in Glacier? 335 days (365 - 30).

Or they might ask: When will the object be deleted? After 365 days. Troubleshooting questions are rare but possible. For example: A user receives an error when trying to download an object directly from Glacier without restoring it first.

The answer is that you cannot read an object directly from Glacier storage; you must first initiate a restore, wait for the copy to be available in S3 Standard, and then download. Another pattern uses retrieval costs: An object stored in Glacier for 45 days is deleted. How much will the customer be charged?

They will be charged an early deletion fee because the minimum storage duration is 90 days. So, the exam will test your knowledge of the 90-day commitment. In Azure exams (AZ-104), while Glacier is not directly used, you may see similar questions about Azure Blob Archive tier, which also has a delay in retrieval (with rehydration times).

The same principles apply. For Google Cloud exams, you might compare Coldline vs Archive vs Standard storage classes, and the logic is identical. Always remember: if data must be instantly available (milliseconds), it cannot be Glacier or any cold storage class.

If cost is the top priority and access is very rare, Glacier or Deep Archive is correct.

S3 Glacier is not just a place to dump files; it requires planning and ongoing management. As an IT professional, you need to understand how lifecycle policies work in detail. A lifecycle policy is a set of rules that define when objects transition between storage classes and when they expire.

Each rule can have multiple transitions and a single expiration action. The transitions happen once per day, so if you set a transition to Glacier after 30 days, the object moves to Glacier sometime after the 30th day, not exactly at midnight. This is fine for most use cases.

You must also be aware of cost implications. For example, transitioning from S3 Standard to Glacier after 30 days means you pay Standard pricing for 30 days, then Glacier pricing thereafter. There is also a lifecycle transition request fee per object.

If you have millions of small objects, these request fees can add up. In practice, many organizations combine objects into larger archives before moving them to Glacier to reduce the number of objects. Another practical consideration is encryption.

By default, S3 Glacier encrypts data at rest using SSE-S3. If you need to use your own keys (SSE-KMS), you can. However, remember that KMS has request costs, and each retrieval will incur KMS API calls.

This can significantly increase cost if you restore many small objects frequently. Monitoring is also key. Set up CloudWatch metrics for S3 buckets, such as 'NumberOfObjects' and 'BucketSizeBytes' per storage class.

This helps you track how much data is in Glacier and forecast costs. Use S3 Inventory to get detailed reports of object metadata, including storage class, last modified date, and encryption status. You can query this inventory with Amazon Athena for advanced analytics.

Finally, always test your restore process. Set up a non-production environment, upload a few files to Glacier, initiate a restore, and measure how long it actually takes. This ensures your service level agreements (SLAs) are realistic.

In production, consider using S3 Event Notifications with AWS Lambda to automate notifications when a restore completes. For example, send an email to the user who requested the file. This creates a seamless user experience despite the delay.

Amazon S3 Glacier is a family of storage classes within Amazon Simple Storage Service (S3) designed for long-term data archiving and backup. The Glacier family includes three distinct classes: S3 Glacier Instant Retrieval, S3 Glacier Flexible Retrieval (formerly simply S3 Glacier), and S3 Glacier Deep Archive. Each class is optimized for a different balance of access frequency, retrieval time, and cost. Understanding the differences is critical for architects and developers selecting the right class for compliance, regulatory, or cost optimization needs.

S3 Glacier Instant Retrieval is the fastest of the Glacier classes, offering retrieval times in the range of milliseconds. It is designed for data that is accessed once a quarter or less but still requires immediate availability when needed. This class is ideal for long-lived, rarely accessed data such as medical records, financial archives, or media assets that may be needed on short notice. The storage cost is higher than Glacier Flexible Retrieval but lower than Standard-IA. The retrieval cost is also higher, reflecting the instant access capability. For exams like the AWS Certified Solutions Architect Associate, you need to know that Instant Retrieval is the only Glacier class with sub-minute retrieval and that it still uses the same S3 API operations (like RestoreObject) but with a different underlying cost structure.

S3 Glacier Flexible Retrieval is the original Glacier class, with retrieval times configurable from 1 minute to 12 hours. The default expedited retrieval returns data in 1 to 5 minutes, but you can also use standard (3-5 hours) or bulk (5-12 hours) retrievals for lower cost. This class is best for backup data that may be needed for disaster recovery or for archives where retrieval speed is not critical but still manageable. The storage cost is lower than Instant Retrieval, but the retrieval cost increases with speed. A common exam scenario is choosing Flexible Retrieval for data that must be restored within a few hours during a disaster, while using Deep Archive for data that can wait up to 48 hours.

S3 Glacier Deep Archive is the most economical storage class in S3, designed for data that is accessed only once or twice a year and can tolerate retrieval times of 12 to 48 hours. This class is ideal for regulatory archives, email archives, or historical data that must be retained for years but rarely accessed. The storage cost is the lowest across all S3 classes, but the minimum storage duration is 180 days (compared to 90 days for Flexible Retrieval and 30 days for Instant Retrieval). Deep Archive also has a per-GB retrieval fee, which is higher than Flexible Retrieval but still low for infrequent access. In exams, you may be asked to recommend Deep Archive when the access pattern is once per year and retrieval within 48 hours is acceptable, or to identify that Deep Archive cannot be used for data that may need immediate access.

All three Glacier classes are configurable through S3 Lifecycle policies, which can transition objects from Standard to Standard-IA, then to Glacier classes based on age. Understanding the transition windows (e.g., minimum 30 days for Instant Retrieval, 90 days for Flexible Retrieval, 180 days for Deep Archive) is a common exam topic. You must also know that objects stored in S3 Glacier are not directly readable via GET; they must first be restored using the RestoreObject API, which copies the data to a temporary location for a specified duration. The restoration process incurs both data retrieval charges and the cost of temporary storage. This is frequently tested in the context of cost optimization and data lifecycle management.

Finally, all Glacier classes support server-side encryption (SSE-S3, SSE-KMS, SSE-C) and S3 Object Lock for Write Once, Read Many (WORM) compliance. Data durability is the same 99.999999999% (11 nines) across all S3 classes. When answering exam questions, pay attention to the retrieval speed requirement and minimum storage duration to select the correct Glacier class. For example, if a question specifies that data must be available within minutes after a restore request, Glacier Instant Retrieval is the only correct choice. If the question mentions a 40-hour retrieval time and lowest cost, Deep Archive is appropriate.

Lifecycle policies in Amazon S3 are a powerful mechanism to automate the transition of objects between storage classes, including the Glacier family, based on time or other rules. For S3 Glacier, lifecycle policies are essential for optimizing costs by moving data that is no longer frequently accessed to colder tiers. The policies can be defined as transition actions or expiration actions. Transition actions move objects from one storage class to another, while expiration actions delete objects after a specified number of days. Understanding the constraints and behaviors of these policies is critical for exam success.

A common lifecycle policy pattern is to transition objects from S3 Standard to S3 Standard-IA after 30 days, then to S3 Glacier Instant Retrieval after 60 days, then to S3 Glacier Flexible Retrieval after 120 days, and finally to S3 Glacier Deep Archive after 180 days. However, there are minimum age requirements for transitions. For example, objects cannot be transitioned from Standard to Standard-IA within the first 30 days. Similarly, transition from Standard-IA to Glacier classes requires the object to be at least 30 days old for Glacier Instant Retrieval, 30 days for Glacier Flexible Retrieval (but the object must be at least 30 days old from the last modification, and the transition must occur at least 30 days after the object was created), and 180 days for Deep Archive. The minimum storage duration charges also apply: if you delete an object within the minimum duration (e.g., 180 days for Deep Archive), you are charged for the full duration. Exams often test your ability to calculate the cost impact of early deletion.

Another important aspect is the use of lifecycle policies with versioning. When a bucket has versioning enabled, lifecycle policies apply to both current and noncurrent versions. You can set separate transitions for noncurrent versions, for example moving noncurrent versions to Glacier after 30 days and deleting them after 365 days. This is commonly used to manage deleted objects that become noncurrent versions. In exam scenarios, you may be asked to design a policy that archives deleted objects to Glacier while keeping current versions in Standard for immediate access. The key is to use the NoncurrentVersionTransition action.

Lifecycle policies also support date-based rules, allowing you to specify a literal date (e.g., January 1, 2025) for transitions or expirations. However, most exams focus on day-based rules because they are more common. You must also know that lifecycle policies are evaluated once per day and that transitions are not instantaneous. If you need to force an immediate transition, you must manually change the storage class via the S3 Console, CLI, or SDK. Understanding the asynchronicity of lifecycle transitions is a frequent exam trick: a question might suggest that a transition happens immediately, but the correct answer acknowledges the daily evaluation cycle.

lifecycle policies can be used in conjunction with S3 Object Lock. If Object Lock is enabled, lifecycle expiration actions can still delete objects, but they must comply with retention settings. For example, if an object has a retention period of 90 days, a lifecycle rule that expires the object after 30 days will not take effect until the retention period ends. This interaction is tested in scenarios involving regulatory compliance where data must be retained for a minimum period.

Finally, note that lifecycle policies apply only to objects stored in S3. They cannot move objects from S3 Glacier back to Standard automatically. Restoring an object from Glacier creates a temporary copy in the Standard class, but the original object remains in Glacier. To permanently move an object out of Glacier, you must copy it to a new object in the desired class. This is a common misconception: many assume lifecycle policies can upgrade storage classes, but they only downgrade colder. For exam preparation, focus on the age constraints, versioning behavior, and interaction with Object Lock. Always check the minimum days for each transition and whether the question involves current or noncurrent versions.

Retrieving data from S3 Glacier is fundamentally different from reading data from other S3 storage classes. While objects in Standard or Standard-IA can be directly accessed using a GET request, objects in Glacier classes are archived and must first be restored to a temporary staging area in the Standard class before they can be read. This restore operation is performed via the RestoreObject API, which initiates an asynchronous job that makes a copy of the object available for a specified number of days. Understanding the restore process, cost implications, and available options is crucial for both exam questions and real-world cost management.

The restore operation allows you to specify the number of days the restored copy should be available (1 to 365 days). During this period, you are charged for both the original Glacier storage and the temporary Standard storage. Retrieval charges apply based on the amount of data restored. For S3 Glacier Flexible Retrieval, you can choose from three retrieval tiers: Expedited (1-5 minutes), Standard (3-5 hours), and Bulk (5-12 hours). Each tier has a different per-GB retrieval cost, with Expedited being the most expensive. For S3 Glacier Deep Archive, only Standard (12 hours) and Bulk (48 hours) tiers are available; there is no Expedited tier. S3 Glacier Instant Retrieval does not require restore because it provides millisecond access, but it has higher storage costs and a per-request retrieval charge.

Costs associated with restore operations include data retrieval fees (per GB restored), data transfer out fees (if you read the restored data to the internet), and temporary storage charges for the restored copy. Data retrieval fees vary by Glacier class and retrieval tier. For example, retrieving 1 TB of data from Glacier Flexible Retrieval using the Standard tier will incur a retrieval fee, while the same retrieval using the Bulk tier costs less per GB. Exams frequently ask you to compare the total cost of retrieval with different tiers or to recommend a tier based on urgency and budget. A common scenario: a disaster recovery test that requires data within 2 hours would need Expedited retrieval, but if the test can wait 5 hours, Standard is cheaper.

A critical exam detail is that restore operations are not free. Even if you use the Bulk tier, you still pay for data retrieval and temporary storage. Also, the restored copy counts against your bucket's quota and can be deleted after the specified days, but you are charged for the entire duration you set. If you delete the restored copy before the end date, you still pay for the full period. Restoring the same object multiple times incurs multiple retrieval charges. This is important in scenarios where data is frequently restored for audit purposes; you might consider moving it to a warmer class instead.

Another important concept is that restore requests have a limited number of concurrent jobs per account. By default, you can have up to 10 expedited retrievals in progress at once, though this can be increased by contacting AWS Support. For Bulk retrievals, there is no limit, but they are processed at a lower priority. In exam questions, you might be asked to identify why a restore operation is taking longer than expected, and the answer could be the concurrent request limit or that the account has a low service quota.

Finally, the restore operation can be automated using S3 Batch Operations, which can restore millions of objects in a single job. This is cost-effective for large archives. You can use S3 Object Lambda to transform data during restore, though that is more advanced. For the exam, remember that restoring data from Glacier is an asynchronous, billable operation with tier-specific speeds and costs. Always check the retrieval speed requirement and the cost tolerance before selecting a tier. The key takeaway: if your application requires immediate access, use Glacier Instant Retrieval or keep a copy in Standard; if you can wait hours, use Flexible Retrieval; if you can wait a day or two, Deep Archive is the cheapest.

Amazon S3 Glacier storage classes are often used to store data that must be retained for regulatory, legal, or compliance reasons. To meet these requirements, S3 Object Lock can be enabled on a bucket that contains Glacier objects, providing Write Once, Read Many (WORM) protection. Object Lock ensures that objects cannot be overwritten or deleted for a specified retention period, either fixed (retention mode) or indefinite (legal hold). Understanding how Object Lock interacts with Glacier classes is essential for exam scenarios involving data immutability, compliance, and audit trails.

Object Lock offers two retention modes: Governance mode and Compliance mode. In Governance mode, most users cannot delete an object, but users with special permissions (like the s3:BypassGovernanceRetention permission) can. In Compliance mode, no one can delete the object, not even the root user, until the retention period expires. Both modes can be applied to objects stored in any Glacier class. For exams, a common question is to recommend Compliance mode for data that must not be altered under any circumstances, such as financial transaction logs or medical records. Governance mode is suitable for situations where an administrator may need to override the lock for legal reasons.

When using Object Lock with Glacier, the retention period starts when the object is created. If you transition an object from Standard to Glacier via a lifecycle policy, the retention period from the original object's creation date applies. If you apply Object Lock to the Glacier object itself (by putting the object directly into Glacier), the retention period starts from the upload date. A critical exam detail is that lifecycle expiration actions cannot delete an object that is under Object Lock until the retention period expires. This means if you have a compliance rule that requires a 10-year retention, a lifecycle policy set to delete after 30 days will be ignored. This is frequently tested in scenarios where a company must retain data for a minimum period but also wants to automate deletion after that period.

Another important aspect is that Object Lock can only be enabled when a bucket is created; it cannot be enabled on an existing bucket. This is a common trick in exam questions: an administrator wants to add WORM protection to an existing bucket but must create a new bucket and migrate the data. Object Lock is a bucket-level setting, but retention settings can be applied per object. When you restore an object from Glacier to Standard, the restored copy does not inherit the Object Lock settings of the original object unless you explicitly apply them. This means you could unintentionally delete a restored copy even if the Glacier original is locked. Understanding this nuance can appear in troubleshooting questions.

For security, Glacier classes support the same encryption options as other S3 classes: SSE-S3 (Amazon S3-managed keys), SSE-KMS (AWS Key Management Service), and SSE-C (customer-provided keys). SSE-KMS is often required for compliance with frameworks like HIPAA or FedRAMP. When you restore an object from Glacier, the restored copy can use a different encryption key if you specify it during the restore request. This can cause confusion if the downstream application expects the original key. In exams, you might be asked to identify that the restored object must be encrypted with the same key for successful decryption.

Finally, S3 Glacier also integrates with AWS CloudTrail for logging all restore operations, bucket-level actions, and Object Lock modifications. This audit trail is crucial for demonstrating compliance during audits. For the AWS Certified Solutions Architect Associate and Security Specialty exams, you should know how to configure CloudTrail to log all S3 events, including those for Glacier. The combination of Object Lock, encryption, and logging makes Glacier a solid choice for regulated industries. When answering questions, focus on the retention mode selection, the interaction with lifecycle policies, and the pre-requisite of enabling Object Lock at bucket creation.