Cost Control
Part of Cloud storage and backups
Choosing storage tiers from access patterns
Use read frequency, retrieval time and retention to choose object-storage tiers, then check total lifecycle and recovery costs.
Classify each data group by expected reads, rewrites, retention and required retrieval time. Match those constraints to named storage tiers, then estimate storage, request, retrieval, transition and minimum-duration costs. Rarely accessed data does not automatically belong in an offline archive: an urgent restore or short retention period may rule one out.
Describe each data group
Separate live assets, older reports and backup history before applying a tier rule. For each group, record:
- Stored volume, growth, object count and typical object size.
- Reads, writes and replacements in normal and busy periods.
- The volume an urgent event might retrieve.
- The longest acceptable wait before data can be used.
- Expected time before deletion or another transition.
Use observed logs or bills where available; label forecasts as assumptions. Include rewrites and class changes, not just reads. Cost a large recovery separately: an average month can conceal it.
Check the actual class or tier
Choose among named tiers by checking both the access pattern and the time readers can wait for data.
| Service example | Possible fit | Limitation to check |
|---|---|---|
| Amazon S3 Standard | Frequently accessed, performance-sensitive objects | The default S3 storage class. |
| Amazon S3 Intelligent-Tiering | Unknown, changing or unpredictable access patterns | A monthly object monitoring and automation charge applies. It moves objects after periods without access; optional Archive Access and Deep Archive Access tiers require restoration. |
| Amazon S3 Standard-IA and S3 One Zone-IA | Infrequently accessed objects that still need millisecond access | Retrieval fees and a 30-day minimum duration apply; objects smaller than 128 KB are billed as 128 KB. |
| Amazon S3 Glacier Instant Retrieval | Archive data that needs immediate access | Objects are immediately available using GET. |
| Amazon S3 Glacier Flexible Retrieval | Rarely accessed, long-lived data that can wait for restoration | A restore is required before reading. Standard retrieval typically takes 3–5 hours; Expedited retrieval for objects under 250 MB typically takes 1–5 minutes. |
| Amazon S3 Glacier Deep Archive | Long-term archive where retrieval can take hours | A restore is required before reading; Standard retrieval typically takes 12 hours. |
| Google Cloud Storage Standard | Frequently accessed objects | No minimum storage duration or retrieval fees. |
| Google Cloud Storage Nearline | Infrequently accessed objects | Retrieval fees and a 30-day minimum storage duration apply. |
| Google Cloud Storage Coldline | Rarely accessed objects | Retrieval fees and a 90-day minimum storage duration apply. |
| Google Cloud Storage Archive | Rarely accessed data that still needs an online access path | Retrieval fees and a 365-day minimum storage duration apply; access does not require offline rehydration. |
| Azure Blob hot | Frequently accessed or modified data | Highest storage costs and lowest access costs among the listed Azure tiers. |
| Azure Blob cool and cold | Less-used block blobs that must remain online | Online access is immediate. Cool has a 30-day minimum duration; cold has a 90-day minimum. Both have higher access costs than warmer tiers. |
| Azure Blob archive | Rarely accessed block blobs whose readers can wait | Offline access, a 180-day minimum duration and retrieval on the order of hours. |
S3 Intelligent-Tiering moves objects to Infrequent Access after 30 consecutive days without access and to Archive Instant Access after 90 days. Optional Archive Access and Deep Archive Access can be added after minimum no-access periods of 90 and 180 days respectively; Standard retrieval typically takes 3–5 hours and 12 hours for those archive tiers. Standard retrievals are free for objects in S3 Intelligent-Tiering.
Google Cloud Storage Autoclass and Azure Smart tier can manage transitions based on usage patterns. Google Cloud Storage class changes can also use a rewrite or Object Lifecycle Management. Compare each option’s retrieval path with the required recovery time: Google Cloud Storage Archive has no offline retrieval step, while Azure Blob archive is offline.
Storage Tier Comparison by Access Pattern and Retrieval Time
- Amazon S3 Standard
- Frequently accessed, performance-sensitive objects
- Amazon S3 Intelligent-Tiering
- Unpredictable access patterns; monthly monitoring fee
- Amazon S3 Standard-IA / One Zone-IA
- Infrequent access; 30-day minimum duration
- Amazon S3 Glacier Instant Retrieval
- Immediate access for archived data
- Amazon S3 Glacier Flexible Retrieval
- 3–5 hours retrieval (standard); 1–5 minutes (expedited)
- Amazon S3 Glacier Deep Archive
- 12-hour retrieval; long-term archive
- Google Cloud Storage Standard
- Frequently accessed; no minimum duration
- Google Cloud Storage Nearline
- Infrequent access; 30-day minimum
- Google Cloud Storage Coldline
- Rarely accessed; 90-day minimum
- Google Cloud Storage Archive
- Archive with online access path; 365-day minimum
- Azure Blob Hot
- High-frequency access; highest storage cost
- Azure Blob Cool
- Less-used blobs; 30-day minimum
- Azure Blob Cold
- Low usage; 90-day minimum
- Azure Blob Archive
- Offline access; 180-day minimum
Calculate before automating
For each viable tier, estimate capacity and time stored, reads, writes and rewrites, request counts, retrieved volume, lifecycle transitions, minimum-duration effects, monitoring charges and applicable data transfer. Multiply expected billable quantities by current rates for the intended location and contract. Compare normal use, a busy period and an urgent restore.
Include transition-request costs and any minimum-duration charges when costing lifecycle rules. S3 Lifecycle can transition or expire objects, and transition requests have costs; expiration can also incur costs for objects in a class with a minimum storage duration. Review rule scope and effects before enabling one: S3 Lifecycle rules apply to existing objects as well as new ones.
Select the least costly candidate that meets the required access time and retention pattern. Revisit the decision when behaviour changes, and include the required recovery time when choosing a tier for backup history.
Key Cost and Duration Constraints Across Tiers
- Minimum Duration: S3 Standard-IA30 days
- Minimum Duration: Google Nearline30 days
- Minimum Duration: Google Coldline90 days
- Minimum Duration: Azure Cool30 days
- Minimum Duration: Azure Cold90 days
- Minimum Duration: Azure Archive180 days
- Retrieval Time: Glacier Flexible Retrieval (Standard)3–5 hours
- Retrieval Time: Glacier Deep Archive12 hours
- Retrieval Time: Azure ArchiveHours (offline)
