# Storage tiering

## What is storage tiering?

Storage tiering is the practice of moving data between different storage classes based on how frequently it is accessed. Cloud providers offer multiple storage tiers at different price points, with faster, more accessible storage costing more, and slower, infrequently accessed storage costing significantly less.

The goal is to match the cost of storing data to its actual value and access patterns, rather than keeping everything on the most expensive tier by default.

## Why storage tiering matters

Storage costs accumulate quietly. Data written to a bucket or volume tends to stay there indefinitely unless someone actively manages it. Over time, large volumes of data that are rarely or never accessed continue to be billed at the same rate as data that is read constantly.

Tiering addresses this by moving data down to cheaper storage classes as it ages or as access frequency drops, without deleting it.

## Storage tiers across AWS, Azure, and GCP

Each provider offers multiple storage classes with different cost and access trade-offs:

**AWS S3**

| Tier | Best for | Retrieval |
|---|---|---|
| S3 Standard | Frequently accessed data | Immediate |
| S3 Intelligent-Tiering | Unknown or changing access patterns | Immediate |
| S3 Standard-IA | Infrequent access, rapid retrieval needed | Immediate, retrieval fee applies |
| S3 Glacier Instant Retrieval | Archive, occasional access | Milliseconds |
| S3 Glacier Deep Archive | Long-term archive, rarely accessed | Hours |

**Azure Blob Storage**

| Tier | Best for | Retrieval |
|---|---|---|
| Hot | Frequently accessed data | Immediate |
| Cool | Data accessed less than once a month | Immediate, retrieval fee applies |
| Cold | Data accessed less than once a quarter | Immediate, retrieval fee applies |
| Archive | Rarely accessed, long-term retention | Hours |

**GCP Cloud Storage**

| Tier | Best for | Retrieval |
|---|---|---|
| Standard | Frequently accessed data | Immediate |
| Nearline | Data accessed less than once a month | Immediate, retrieval fee applies |
| Coldline | Data accessed less than once a quarter | Immediate, retrieval fee applies |
| Archive | Long-term retention, rarely accessed | Milliseconds |

## How tiering works in practice

Most organizations use life cycle policies to automate tiering rather than managing it manually. A life cycle policy defines rules that move data between storage classes based on age, access frequency, or both. Once configured, the transitions happen automatically without any ongoing intervention.

A typical policy for application logs or backup data might look like this:

- Data stays in Standard storage for the first 30 days, when it is most likely to be accessed for debugging or review.
- After 30 days, it moves to an infrequent access tier where storage costs drop but retrieval remains fast if needed.
- After 90 days, it transitions to archive storage where costs are minimal. On AWS and Azure, retrieval slows to hours instead of milliseconds—GCP keeps millisecond retrieval at every tier.
- After 365 days, it is deleted automatically if there is no compliance or retention requirement to keep it longer.

The main benefit beyond cost savings is that tiering becomes a set-and-forget process. Once life cycle policies are in place, data moves down through tiers automatically as it ages, without relying on anyone to remember to do it.

## An example of storage tiering

Say a company stores application logs in AWS S3 Standard. Logs older than 30 days are rarely accessed but retained for compliance. The table below outlines storage tiering costs with a life cycle policy in place.

| Data age | Storage class | Monthly cost per TB |
|---|---|---|
| 0 to 30 days | S3 Standard | $23 |
| 30 to 90 days | S3 Standard-IA | $12.50 |
| 90 days+ | S3 Glacier Deep Archive | $0.99 |

If the organization retains 100TB of logs, there's substantial savings from tiering older data compared to keeping everything in S3 Standard.

Note: Storage class names and pricing are subject to change. Check the AWS, Azure, and GCP pricing pages for current rates.

## What to watch out for

- Retrieval fees apply on lower tiers. Moving data to archive storage is cheap, but retrieving it can cost more than keeping it on a mid-tier. Factor retrieval frequency into the decision before committing to a lower tier.
- Minimum storage duration charges apply on most tiers below Standard. Moving data to Glacier and then retrieving it after a few days can end up costing more than leaving it on Standard-IA.
- Intelligent-Tiering, which moves data automatically based on access patterns, adds a small per-object monitoring fee. For buckets with large numbers of small objects, this fee can exceed the tiering savings.