Module Navigation

Previous: Compute | Parent: Storage Learning Path | Next: Object vs Block vs File Storage QA

---

Introduction

Cloud storage is a managed service that allows applications, users, and systems to store data on infrastructure operated by a cloud provider.

Instead of purchasing, configuring, and maintaining physical storage devices, organizations consume storage as an on-demand service.

Cloud storage is used for:

• Application data
• Virtual-machine disks
• Database files
• Documents
• Images and videos
• Backups
• Log files
• Data lakes
• Machine-learning datasets
• Shared file systems
• Disaster-recovery copies
• Archived records

The three major cloud-storage models are:

• Object storage
• Block storage
• File storage

Each model solves a different problem.

Choosing the wrong storage type can lead to:

• Poor performance
• High cost
• Data-loss risk
• Scalability limitations
• Complicated operations
• Security issues

What Is Cloud Storage?

Cloud storage is a service that stores data on remotely managed infrastructure and makes it accessible through APIs, network protocols, or attached storage volumes.

The cloud provider generally manages:

• Physical disks
• Storage servers
• Hardware replacement
• Replication
• Capacity expansion
• Data durability
• Infrastructure monitoring
• Platform availability

The customer remains responsible for:

• Choosing the correct storage type
• Configuring permissions
• Classifying data
• Encrypting sensitive information
• Defining lifecycle policies
• Configuring backup and recovery
• Monitoring usage
• Controlling cost
• Meeting compliance requirements

---

Basic Cloud Storage Architecture

flowchart TB
    Users[Users and Applications] --> Access[API or Storage Protocol]

    Access --> StorageService[Cloud Storage Service]

    StorageService --> Metadata[Metadata Layer]
    StorageService --> DataLayer[Distributed Data Layer]

    DataLayer --> NodeA[Storage Node A]
    DataLayer --> NodeB[Storage Node B]
    DataLayer --> NodeC[Storage Node C]

    NodeA --> Replication[Replication and Redundancy]
    NodeB --> Replication
    NodeC --> Replication

    Replication --> Monitoring[Health Monitoring]

Cloud storage platforms distribute data across multiple physical devices and often across multiple failure domains.

---

Why Cloud Storage Is Important

Traditional storage requires organizations to estimate future capacity and purchase hardware in advance.

Cloud storage provides:

• On-demand capacity
• Elastic scaling
• Pay-as-you-use pricing
• Managed durability
• Managed availability
• Global accessibility
• API-based automation
• Lifecycle management
• Encryption integration
• Backup integration
• Monitoring integration

This allows teams to focus more on application design and less on storage-hardware administration.

---

Major Cloud Storage Types

The three major storage types are:

Storage Type	Data Organization	Common Use
Object storage	Objects in buckets or containers	Images, backups, logs, data lakes
Block storage	Fixed-size storage blocks	VM disks, databases, operating systems
File storage	Files and directories	Shared folders, enterprise applications

---

Object Storage

Object storage stores data as independent objects.

Each object usually contains:

• Object data
• Metadata
• Unique object key

Objects are stored inside logical containers such as:

• Buckets
• Containers

Examples include:

• Amazon S3
• Azure Blob Storage
• Google Cloud Storage

flowchart LR
    Application[Application] --> API[REST API]
    API --> Bucket[Storage Bucket]

    Bucket --> ObjectA[Object A]
    Bucket --> ObjectB[Object B]
    Bucket --> ObjectC[Object C]

Object storage is suitable for:

• Images
• Videos
• Documents
• Backups
• Static website files
• Logs
• Data lakes
• Archived records

---

Block Storage

Block storage divides storage capacity into fixed-size blocks.

The operating system sees block storage as a disk device.

A file system or database can then be created on top of the volume.

Examples include:

• Amazon EBS
• Azure Managed Disks
• Google Persistent Disk

flowchart LR
    VirtualMachine[Virtual Machine] --> Volume[Block Storage Volume]
    Volume --> Block1[Block 1]
    Volume --> Block2[Block 2]
    Volume --> Block3[Block 3]

Block storage is suitable for:

• Operating-system disks
• Relational databases
• Transactional applications
• Virtual-machine storage
• Applications requiring low-latency disk access

---

File Storage

File storage organizes data using files and directories.

It is commonly accessed using protocols such as:

• NFS
• SMB

Examples include:

• Amazon EFS
• Amazon FSx
• Azure Files
• Google Filestore

flowchart LR
    AppA[Application Server A] --> SharedFS[Shared File System]
    AppB[Application Server B] --> SharedFS
    AppC[Application Server C] --> SharedFS

    SharedFS --> DirectoryA[/Reports]
    SharedFS --> DirectoryB[/Documents]
    SharedFS --> DirectoryC[/Uploads]

File storage is suitable for:

• Shared application files
• Content-management systems
• User home directories
• Media workflows
• Legacy enterprise applications
• Shared configuration files

---

Persistent vs Ephemeral Storage

Persistent Storage

Persistent storage retains data after:

• Application restart
• Virtual-machine restart
• Container restart
• Function execution completion

Examples include:

• Object storage
• Persistent disks
• Managed file systems
• Databases

Ephemeral Storage

Ephemeral storage exists only for the lifetime of a compute instance, container, or function environment.

Examples include:

• Local VM temporary disk
• Container writable layer
• Function temporary directory
• Instance-local SSD

flowchart TD
    Compute[Compute Instance] --> Persistent[Persistent Storage]
    Compute --> Ephemeral[Ephemeral Storage]

    Persistent -->|Compute Deleted| Retained[Data Retained]
    Ephemeral -->|Compute Deleted| Lost[Data May Be Lost]

Ephemeral storage should not be used for critical persistent data.

---

Local Storage vs Network Storage

Local Storage

Local storage is physically attached to the compute host.

Advantages:

• Low latency
• High throughput
• No network dependency

Limitations:

• Data may be lost when the instance stops or fails
• Difficult to share between systems
• Limited portability

Network Storage

Network storage is accessed over a network.

Advantages:

• Persistent
• Easier to scale
• Can support replication
• May be shared
• Independent from the compute host

Limitations:

• Network latency
• Potential network bottlenecks
• Additional service cost

---

Storage Performance Metrics

Cloud-storage performance is commonly measured using:

• Latency
• IOPS
• Throughput
• Bandwidth
• Request rate
• Queue depth
• Capacity

---

Latency

Latency is the time required to complete one storage operation.

Examples include:

• Reading a database block
• Writing a file
• Retrieving an object
• Updating metadata

Latency is commonly measured in:

• Microseconds
• Milliseconds

flowchart LR
    Request[Storage Request] --> Storage[Storage System]
    Storage --> Response[Storage Response]

    Request -. Elapsed Time .-> Response

Low latency is especially important for:

• Databases
• Transaction processing
• Interactive applications
• Operating-system disks

---

IOPS

IOPS means Input/Output Operations Per Second.

It measures how many read and write operations a storage system can complete per second.

Examples:

• 3,000 IOPS
• 10,000 IOPS
• 100,000 IOPS

IOPS is important for workloads that perform many small operations.

Examples include:

• Relational databases
• Transaction-processing systems
• Virtual desktops
• Metadata-heavy applications

---

Throughput

Throughput measures how much data can be transferred per unit of time.

It is commonly measured in:

• MB/s
• GB/s

Throughput is important for workloads that process large sequential files.

Examples include:

• Video processing
• Data analytics
• Backup transfer
• Large file processing
• Machine-learning datasets

---

IOPS vs Throughput

IOPS	Throughput
Measures operation count	Measures data volume
Important for small random operations	Important for large sequential transfers
Common for databases	Common for analytics and media
Measured in operations per second	Measured in MB/s or GB/s

A workload can require high IOPS, high throughput, or both.

---

Example Performance Comparison

Consider two workloads.

Workload A: Transaction Database

• Thousands of small reads
• Thousands of small writes
• Random access
• Low-latency requirement

Primary requirement:

• High IOPS
• Low latency

Workload B: Video Processing

• Reads very large files
• Writes transformed files
• Mostly sequential access

Primary requirement:

• High throughput
• High bandwidth

---

Storage Capacity

Capacity is the amount of data that can be stored.

It is commonly measured in:

• GB
• TB
• PB

Cloud storage can generally scale on demand, but each service may have limits involving:

• Volume size
• File size
• Object size
• Number of objects
• Request rate
• Number of mounted clients
• Throughput

Service quotas should be reviewed during architecture design.

---

Durability

Durability measures the probability that stored data will remain intact over time.

A highly durable storage service protects data from:

• Disk failure
• Server failure
• Rack failure
• Data corruption
• Hardware replacement
• Some facility-level failures

Durability is improved through:

• Replication
• Erasure coding
• Integrity checks
• Automatic repair
• Redundant hardware
• Geographic copies

---

Availability

Availability measures whether the storage service can be accessed when required.

A storage service may be highly durable but temporarily unavailable.

flowchart TD
    Data[Stored Data] --> Durable{Data Preserved?}
    Data --> Available{Accessible Now?}

    Durable -->|Yes| Safe[Data Is Safe]
    Available -->|No| Outage[Temporary Access Issue]

Durability and availability solve different problems.

---

Durability vs Availability

Durability	Availability
Measures data survival	Measures service accessibility
Protects against data loss	Protects against downtime
Improved through replication	Improved through redundant service endpoints
Long-term property	Operational property

Interview candidates should not use these terms interchangeably.

---

Redundancy

Redundancy means storing extra copies or coded fragments of data.

Storage redundancy may exist across:

• Multiple disks
• Multiple servers
• Multiple racks
• Multiple availability zones
• Multiple regions

flowchart LR
    Original[Original Data] --> CopyA[Copy in Zone A]
    Original --> CopyB[Copy in Zone B]
    Original --> CopyC[Copy in Zone C]

More redundancy generally improves resilience but may increase cost.

---

Replication

Replication creates copies of data.

Common replication models include:

• Synchronous replication
• Asynchronous replication
• Same-zone replication
• Cross-zone replication
• Cross-region replication

---

Synchronous Replication

With synchronous replication, the write is confirmed after required replicas acknowledge it.

sequenceDiagram
    participant App
    participant Primary
    participant Replica
    App->>Primary: Write data
    Primary->>Replica: Replicate data
    Replica-->>Primary: Replication complete
    Primary-->>App: Write successful

Advantages:

• Stronger consistency
• Lower data-loss risk

Limitations:

• Higher write latency
• Dependency on replica availability

---

Asynchronous Replication

With asynchronous replication, the primary system acknowledges the write before all replicas are updated.

sequenceDiagram
    participant App
    participant Primary
    participant Replica
    App->>Primary: Write data
    Primary-->>App: Write successful
    Primary->>Replica: Replicate later

Advantages:

• Lower write latency
• Better geographic flexibility

Limitations:

• Replication lag
• Possible data loss during immediate failure

---

Availability Zones and Regions

Availability Zone

An availability zone is an isolated location within a cloud region.

A multi-zone storage service can survive some zone-level failures.

Region

A region is a geographic area containing multiple cloud facilities or availability zones.

Cross-region storage supports:

• Disaster recovery
• Geographic access
• Compliance
• Regional failover

---

Storage Consistency

Consistency describes when updated data becomes visible to readers.

Common consistency models include:

• Strong consistency
• Eventual consistency

---

Strong Consistency

After a successful write, all later reads return the latest value.

sequenceDiagram
    participant Writer
    participant Storage
    participant Reader
    Writer->>Storage: Write version 2
    Storage-->>Writer: Success
    Reader->>Storage: Read object
    Storage-->>Reader: Version 2

Strong consistency simplifies application logic.

---

Eventual Consistency

After a write, some readers may temporarily receive an older value.

Eventually, all replicas converge on the latest value.

flowchart LR
    Write[Write New Value] --> ReplicaA[Replica A Updated]
    Write --> Delay[Replication Delay]
    Delay --> ReplicaB[Replica B Updated]

Applications using eventual consistency may need:

• Retry logic
• Version numbers
• Conflict resolution
• Read-after-write handling
• Idempotency

---

Storage Tiers and Classes

Cloud providers commonly offer multiple storage tiers based on access frequency.

Typical tiers include:

• Hot
• Cool
• Cold
• Archive

---

Hot Storage

Hot storage is designed for frequently accessed data.

Characteristics:

• Low retrieval latency
• Higher storage cost
• Lower or no retrieval fee
• Suitable for active application data

Examples:

• Website content
• Active documents
• Frequently used datasets
• Current application logs

---

Cool Storage

Cool storage is designed for infrequently accessed data.

Characteristics:

• Lower storage cost
• Possible retrieval charges
• Minimum retention requirements
• Suitable for older backups and reports

---

Cold Storage

Cold storage is designed for rarely accessed data.

Characteristics:

• Lower storage cost
• Higher retrieval cost
• Possible retrieval delay
• Longer minimum retention period

Use cases include:

• Historical data
• Compliance copies
• Old project data
• Long-term backups

---

Archive Storage

Archive storage is designed for long-term retention.

Characteristics:

• Very low storage cost
• Retrieval may take minutes or hours
• Higher retrieval fees
• Long minimum retention periods

Use cases include:

• Legal records
• Regulatory data
• Long-term backup
• Historical archives

---

Storage-Tier Comparison

Tier	Access Frequency	Retrieval Speed	Storage Cost
Hot	Frequent	Immediate	Highest
Cool	Occasional	Immediate or near-immediate	Lower
Cold	Rare	Slower or immediate depending on service	Low
Archive	Very rare	Minutes to hours	Lowest

The exact names and behavior vary by cloud provider.

---

Storage Lifecycle Management

Lifecycle management automatically moves or deletes data based on rules.

Example lifecycle:

flowchart LR
    Hot[Hot Storage: Day 0] --> Cool[Cool Storage: Day 30]
    Cool --> Cold[Cold Storage: Day 90]
    Cold --> Archive[Archive: Day 365]
    Archive --> Delete[Delete: Year 7]

Lifecycle rules can:

• Move data to cheaper tiers
• Delete expired data
• Remove old versions
• Clean incomplete uploads
• Enforce retention periods
• Reduce manual administration

---

Data Retention

Data retention defines how long information should be stored.

Retention requirements may come from:

• Business policy
• Legal requirements
• Audit requirements
• Regulatory standards
• Customer agreements
• Security policy

Retaining data forever is usually not a good default.

Long retention increases:

• Cost
• Security exposure
• Compliance scope
• Discovery burden
• Privacy risk

---

Backup

A backup is a separate recoverable copy of data.

A backup protects against:

• Accidental deletion
• Corruption
• Application bugs
• Ransomware
• Incorrect updates
• Primary storage failure

flowchart LR
    Primary[(Primary Storage)] --> Backup[(Backup Storage)]
    Backup --> Restore[Restore Process]
    Restore --> Recovered[(Recovered Data)]

Replication alone is not always a backup because accidental changes may be replicated.

---

Snapshot

A snapshot captures the state of storage at a specific point in time.

Snapshots are commonly used for:

• Block volumes
• Virtual machines
• Databases
• File systems

Snapshots may be:

• Full
• Incremental
• Crash-consistent
• Application-consistent

Application-consistent snapshots coordinate with the application before capturing data.

---

Replication vs Backup

Replication	Backup
Maintains copies for availability	Maintains recoverable historical copies
Changes are copied quickly	Copies may be retained over time
Accidental deletion may replicate	Older backup can restore deleted data
Supports failover	Supports recovery
Focuses on uptime	Focuses on data restoration

A resilient system often needs both replication and backup.

---

Recovery Point Objective

Recovery Point Objective, or RPO, defines the maximum acceptable data loss measured in time.

Example:

RPO = 15 minutes

The organization accepts losing no more than 15 minutes of data.

Lower RPO usually requires:

• More frequent backups
• Continuous replication
• Change-data capture
• Higher cost

---

Recovery Time Objective

Recovery Time Objective, or RTO, defines the maximum acceptable recovery duration.

Example:

RTO = 1 hour

The service must be restored within one hour.

flowchart LR
    Failure[Failure Occurs] --> Recovery[Recovery Process]
    Recovery --> Service[Service Restored]

    Failure -. RTO Window .-> Service

---

RPO vs RTO

RPO	RTO
Maximum acceptable data loss	Maximum acceptable downtime
Measures time between data copies	Measures recovery duration
Improved with frequent replication or backup	Improved with automation and standby capacity
Data-focused	Service-focused

---

Encryption at Rest

Encryption at rest protects stored data.

Cloud storage may use:

• Provider-managed keys
• Customer-managed keys
• Customer-supplied keys
• Application-level encryption

flowchart LR
    Application[Application] --> Encrypt[Encryption]
    Encrypt --> Storage[(Encrypted Storage)]
    KeyManager[Key Management Service] --> Encrypt

Customer-managed keys offer greater control but require:

• Key-policy management
• Rotation
• Monitoring
• Backup
• Availability planning

---

Encryption in Transit

Encryption in transit protects data moving between:

• Clients and storage
• Applications and storage
• Replication endpoints
• Regions
• Backup systems

Common protection methods include:

• HTTPS
• TLS
• Secure file-transfer protocols
• Encrypted VPN connections

Unencrypted storage traffic should be avoided.

---

Access Control

Storage access should follow least privilege.

Common access-control mechanisms include:

• IAM roles
• Service accounts
• Access-control lists
• Bucket policies
• File permissions
• Network restrictions
• Signed URLs
• Temporary credentials

flowchart LR
    User[User or Application] --> Identity[Identity]
    Identity --> Policy[Access Policy]
    Policy --> Storage[(Storage Resource)]

Public access should be disabled unless explicitly required.

---

Storage Security Layers

flowchart TB
    User[User or Application] --> Authentication[Authentication]
    Authentication --> Authorization[Authorization]
    Authorization --> Network[Network Controls]
    Network --> Encryption[Encryption]
    Encryption --> Storage[(Cloud Storage)]

    Storage --> Logging[Audit Logging]
    Logging --> Monitoring[Security Monitoring]

A secure storage architecture uses multiple layers rather than relying on one control.

---

Data Classification

Data should be classified based on sensitivity.

Example classification:

Classification	Example	Protection Level
Public	Public website images	Basic integrity controls
Internal	Internal documents	Authenticated access
Confidential	Customer records	Strong encryption and restricted access
Restricted	Payment or health information	Highest controls and auditing

Storage type, encryption, retention, and access should reflect the classification.

---

Storage Naming and Organization

Storage should be organized consistently.

Examples include:

• Separate storage by environment
• Separate production and non-production data
• Use clear naming standards
• Use prefixes or folders by domain
• Apply tags and labels
• Assign ownership
• Document retention requirements

Example:

prod-customer-documents
prod-audit-archive
dev-application-logs
test-data-processing

Do not mix production and development data unless there is a clear approved design.

---

Storage Metadata

Metadata provides information about stored data.

Examples include:

• Content type
• Creation date
• Owner
• Classification
• Retention date
• Checksum
• Version
• Source application

Metadata supports:

• Search
• Governance
• Lifecycle policies
• Audit
• Data processing
• Security classification

---

Checksums and Integrity

A checksum is a calculated value used to detect data corruption.

flowchart LR
    Original[Original File] --> HashA[Calculate Checksum]
    Transfer[Transfer or Store File] --> HashB[Calculate Checksum Again]

    HashA --> Compare{Checksums Match?}
    HashB --> Compare

    Compare -->|Yes| Valid[Data Is Valid]
    Compare -->|No| Corrupt[Possible Corruption]

Storage systems may use checksums to detect and repair corrupted data.

---

Versioning

Versioning keeps multiple versions of an object or file.

It protects against:

• Accidental overwrite
• Accidental deletion
• Application bugs
• Malicious modification

flowchart LR
    Object[report.pdf] --> V1[Version 1]
    Object --> V2[Version 2]
    Object --> V3[Version 3]

Versioning improves recoverability but can increase storage cost.

Lifecycle rules should be used to delete unnecessary old versions.

---

Immutable Storage

Immutable storage prevents data from being modified or deleted during a retention period.

It is useful for:

• Audit logs
• Compliance records
• Financial records
• Ransomware protection
• Legal holds

Immutability may use:

• Write once, read many controls
• Retention locks
• Legal holds
• Protected backup vaults

---

Multi-Tenancy

Cloud storage services often support many customers on shared physical infrastructure.

Isolation is maintained through:

• Identity boundaries
• Encryption
• Logical resource separation
• Network controls
• Platform isolation

Customers must still correctly configure permissions.

Many cloud-storage incidents result from misconfiguration rather than platform isolation failure.

---

Storage Cost Components

Storage cost may include:

• Capacity consumed
• Number of operations
• Data retrieval
• Data transfer
• Replication
• Snapshots
• Backup
• Early deletion
• Provisioned IOPS
• Provisioned throughput

---

Common Storage Operations

Cloud providers may charge for:

• Read operations
• Write operations
• List operations
• Delete operations
• Metadata requests
• Lifecycle transitions
• Retrieval requests

A workload with millions of small operations may have significant request cost even when total capacity is small.

---

Data Transfer Cost

Data transfer may be charged when data moves:

• Between regions
• To the public internet
• Between services
• Across availability zones
• To another cloud provider
• To an on-premises environment

flowchart LR
    RegionA[Cloud Region A] --> RegionB[Cloud Region B]
    RegionA --> Internet[Public Internet]
    RegionA --> OnPrem[On-Premises Data Center]

Data location should be considered during architecture planning.

---

Production Storage Architecture

flowchart TB
    Users[Users] --> CDN[Content Delivery Network]
    CDN --> ObjectStorage[(Object Storage)]

    Application[Application Services] --> BlockStorage[(Block Storage)]
    Application --> FileStorage[(Shared File Storage)]
    Application --> Cache[(Distributed Cache)]

    ObjectStorage --> Replication[Cross-Region Replication]
    BlockStorage --> Snapshots[Volume Snapshots]
    FileStorage --> Backup[Managed Backup]

    Replication --> DR[Disaster-Recovery Region]
    Snapshots --> BackupVault[Backup Vault]
    Backup --> BackupVault

    ObjectStorage --> Lifecycle[Lifecycle Management]
    Lifecycle --> Archive[Archive Storage]

    ObjectStorage --> Monitoring[Monitoring and Audit]
    BlockStorage --> Monitoring
    FileStorage --> Monitoring

---

Production Use Case: Online Document Platform

Consider an online document-management platform.

The application stores several kinds of data:

Data	Storage Choice
Uploaded documents	Object storage
Application database	Block-backed managed database
Shared processing files	File storage
Temporary file transformation	Ephemeral storage
Historical documents	Archive storage
Database recovery points	Snapshots and backups

---

Processing Flow

1. A user uploads a document.
2. The API validates the file type and size.
3. The document is stored in object storage.
4. Metadata is saved in the application database.
5. An event starts virus scanning and content processing.
6. Processed output is stored in a separate object prefix.
7. Older documents move to lower-cost storage using lifecycle rules.
8. Object versioning protects against accidental overwrites.
9. Cross-region replication protects critical documents.
10. Audit logs are stored in immutable storage.
11. Backup policies protect application metadata.
12. Monitoring detects access failures, unusual downloads, and backup problems.

---

Storage Selection Framework

When selecting storage, ask the following questions:

• Is the data structured or unstructured?
• Does the application need a mounted file system?
• Does the operating system need a disk?
• Is low latency required?
• Is the data frequently accessed?
• Must multiple servers access it?
• Is strict consistency required?
• How much throughput is needed?
• How many IOPS are needed?
• What are the RPO and RTO?
• How long should the data be retained?
• Does the data require immutability?
• Must the data exist in multiple regions?
• What is the expected cost?

---

Interview Questions and Answers

1. What is cloud storage?

Answer

Cloud storage is a managed service that stores data on infrastructure operated by a cloud provider.

Applications access the storage through:

• APIs
• Mounted volumes
• File-system protocols
• Storage SDKs

Cloud providers manage the physical infrastructure, capacity, hardware failure, and much of the replication.

Customers remain responsible for storage selection, permissions, encryption, backup, lifecycle, compliance, and cost management.

---

2. What are the main types of cloud storage?

Answer

The three main types are:

• Object storage
• Block storage
• File storage

Object storage is suitable for unstructured data such as images, backups, and logs.

Block storage is suitable for virtual-machine disks and databases.

File storage is suitable for shared directories and applications that require NFS or SMB access.

---

3. What is the difference between persistent and ephemeral storage?

Answer

Persistent storage retains data independently of the compute resource.

Ephemeral storage exists only for the lifetime of a compute instance, container, or function environment.

Persistent storage should be used for critical application data.

Ephemeral storage is appropriate for:

• Temporary files
• Caches
• Intermediate processing
• Downloaded artifacts that can be recreated

---

4. What are IOPS, throughput, and latency?

Answer

IOPS measures the number of read or write operations completed per second.

Throughput measures the amount of data transferred per second.

Latency measures how long one operation takes.

Database workloads often require high IOPS and low latency.

Media and analytics workloads often require high throughput.

---

5. What is the difference between durability and availability?

Answer

Durability measures whether data remains intact over time.

Availability measures whether the data can be accessed when needed.

A service can preserve data safely while being temporarily unavailable.

Durability protects against data loss, while availability protects against service interruption.

---

6. What is replication?

Answer

Replication creates additional copies of data across disks, servers, availability zones, or regions.

Replication improves:

• Availability
• Durability
• Disaster recovery
• Read scalability

Synchronous replication provides stronger consistency but may increase latency.

Asynchronous replication reduces write latency but may create replication lag.

---

7. What is the difference between backup and replication?

Answer

Replication maintains current copies of data to support availability and failover.

Backup maintains recoverable historical copies.

If a user accidentally deletes data, replication may copy the deletion to all replicas.

A backup may allow restoration from a point before the deletion.

Production systems commonly require both replication and backup.

---

8. What are RPO and RTO?

Answer

RPO is the maximum acceptable amount of data loss measured in time.

RTO is the maximum acceptable amount of downtime.

Example:

• RPO of 15 minutes means no more than 15 minutes of data may be lost.
• RTO of one hour means the service should be restored within one hour.

These requirements guide backup, replication, and disaster-recovery design.

---

9. What are storage tiers?

Answer

Storage tiers are pricing and performance classes designed for different access frequencies.

Typical tiers include:

• Hot
• Cool
• Cold
• Archive

Frequently accessed data should remain in hot storage.

Rarely accessed long-term data may be moved to cold or archive storage.

Lifecycle policies can automate these transitions.

---

10. What is storage lifecycle management?

Answer

Storage lifecycle management automatically transitions or deletes data based on configured rules.

For example:

• Keep data in hot storage for 30 days.
• Move it to cool storage after 30 days.
• Move it to archive after one year.
• Delete it after seven years.

Lifecycle management reduces cost and enforces retention policy.

---

11. How do you secure cloud storage?

Answer

Cloud storage should be secured through:

• Least-privilege IAM
• Private access by default
• Encryption at rest
• Encryption in transit
• Network restrictions
• Temporary credentials
• Audit logging
• Versioning
• Backup
• Data classification
• Key-management controls

Sensitive data should never be stored in publicly accessible resources unless explicitly required and protected.

---

12. What is storage versioning?

Answer

Versioning preserves multiple versions of an object or file.

It protects against:

• Accidental overwrite
• Accidental deletion
• Application errors
• Some ransomware events

Versioning increases storage consumption, so lifecycle policies should manage old versions.

---

13. What is immutable storage?

Answer

Immutable storage prevents data from being changed or deleted for a defined retention period.

It is commonly used for:

• Audit logs
• Regulatory records
• Legal evidence
• Backup protection
• Financial records

Immutability helps protect data from accidental or malicious deletion.

---

14. How do you choose the correct cloud-storage type?

Answer

Choose storage based on access pattern and application requirements.

Use object storage for:

• Unstructured data
• Backups
• Logs
• Static content
• Data lakes

Use block storage for:

• VM disks
• Databases
• Low-latency transactional workloads

Use file storage for:

• Shared directories
• NFS or SMB applications
• Legacy enterprise applications

Also consider performance, availability, durability, security, retention, and cost.

---

15. What are common cloud-storage design mistakes?

Answer

Common mistakes include:

• Using ephemeral storage for critical data
• Confusing durability with backup
• Making storage publicly accessible
• Ignoring lifecycle policies
• Selecting storage only by capacity cost
• Ignoring request and transfer charges
• Not configuring versioning
• Not testing restores
• Overlooking RPO and RTO
• Using one storage type for every workload
• Ignoring concurrency and throughput limits
• Storing sensitive data without classification

---

Common Interview Follow-Up Questions

• What is cloud storage?
• How does object storage work?
• What is the difference between a disk and a file share?
• What does eleven nines of durability mean?
• Does replication replace backup?
• What is storage latency?
• What is provisioned IOPS?
• What is cross-region replication?
• What is a storage snapshot?
• What is immutable backup?
• What is read-after-write consistency?
• What is a lifecycle policy?
• How do data-transfer costs affect architecture?
• When should archive storage be used?
• How would you design storage for a multi-region application?

---

Common Mistakes

Using One Storage Type for Everything

Object, block, and file storage serve different access patterns.

Storing Critical Data on Ephemeral Disk

Temporary storage may disappear when compute resources restart or fail.

Assuming Replication Is Backup

Replication can copy accidental deletion and corruption.

Ignoring Restore Testing

A backup is useful only when it can be restored successfully.

Confusing Durability and Availability

Data survival and service accessibility are different concerns.

Making Buckets Public

Public storage can expose confidential information.

Ignoring Storage Request Cost

Millions of small operations can create significant cost.

Ignoring Data-Transfer Charges

Cross-region and internet transfers may be expensive.

Keeping All Data in Hot Storage

Lifecycle policies should move old data to cheaper tiers.

Missing Versioning

Accidental overwrites may become difficult to recover.

No RPO or RTO

Recovery architecture cannot be designed without business targets.

Over-Provisioning Performance

Provisioned IOPS and throughput should match measured workload requirements.

---

Troubleshooting Cloud Storage

Application Cannot Access Storage

Check:

• IAM permissions
• Resource policy
• Network access
• Storage endpoint
• Encryption-key permission
• Account or subscription
• Region
• Resource name

Storage Is Slow

Check:

• Latency
• IOPS limits
• Throughput limits
• Network bandwidth
• Request pattern
• File size
• Concurrency
• Storage tier
• Database query pattern

Unexpectedly High Cost

Check:

• Capacity growth
• Request volume
• Data retrieval
• Data transfer
• Snapshot retention
• Old object versions
• Replication
• Provisioned performance
• Incomplete uploads

Data Was Accidentally Deleted

Check:

• Versioning
• Soft delete
• Snapshot
• Backup
• Replication status
• Retention lock
• Recovery procedure

Replication Is Delayed

Check:

• Replication configuration
• Destination permissions
• Encryption-key permissions
• Regional service health
• Object eligibility
• Replication backlog

Backup Cannot Be Restored

Check:

• Backup integrity
• Encryption keys
• Restore permissions
• Application compatibility
• Snapshot state
• Recovery documentation
• Test environment

---

Cloud Storage Best Practices

• Choose storage based on access pattern.
• Separate production and non-production data.
• Use persistent storage for critical data.
• Treat ephemeral storage as temporary.
• Enable encryption in transit.
• Enable encryption at rest.
• Apply least-privilege access.
• Disable public access by default.
• Classify sensitive data.
• Enable audit logging.
• Use versioning where appropriate.
• Configure lifecycle policies.
• Delete unnecessary old versions.
• Configure replication based on resilience needs.
• Create independent backups.
• Define RPO and RTO.
• Test restore procedures.
• Monitor latency, IOPS, and throughput.
• Monitor capacity growth.
• Monitor request and transfer cost.
• Use immutable storage for critical records.
• Document data ownership.
• Review storage quotas.
• Automate infrastructure configuration.
• Use checksums for integrity validation.

---

Storage Basics Checklist

Architecture

• Identify the required storage model.
• Document access patterns.
• Estimate capacity growth.
• Estimate IOPS and throughput.
• Define consistency requirements.
• Define sharing requirements.

Reliability

• Define RPO.
• Define RTO.
• Configure replication.
• Configure backup.
• Enable versioning.
• Test restoration.
• Consider multi-zone or multi-region storage.

Security

• Classify data.
• Apply least privilege.
• Disable unnecessary public access.
• Encrypt data at rest.
• Encrypt data in transit.
• Restrict network access.
• Enable audit logs.
• Protect encryption keys.

Cost

• Choose the correct storage tier.
• Configure lifecycle rules.
• Delete old versions.
• Review snapshot retention.
• Monitor request volume.
• Review data-transfer charges.
• Review replication costs.

Operations

• Monitor capacity.
• Monitor latency.
• Monitor IOPS.
• Monitor throughput.
• Alert on backup failures.
• Alert on replication failures.
• Document ownership.
• Maintain recovery runbooks.

---

Quick Revision

Topic	Key Point
Cloud storage	Managed remote data-storage service
Object storage	Stores objects in buckets
Block storage	Provides disk-like volumes
File storage	Provides files and directories
Persistent storage	Retains data beyond compute lifetime
Ephemeral storage	Temporary storage tied to compute
Latency	Time required for one operation
IOPS	Operations completed per second
Throughput	Data transferred per second
Durability	Probability of data survival
Availability	Ability to access the service
Replication	Maintains additional data copies
Backup	Stores recoverable historical copies
Snapshot	Point-in-time storage copy
RPO	Maximum acceptable data loss
RTO	Maximum acceptable downtime
Lifecycle	Automates transition and deletion
Versioning	Preserves previous data versions
Immutability	Prevents modification or deletion
Encryption	Protects stored and transferred data

---

Key Takeaways

• Cloud storage provides managed, scalable, and durable data storage.
• Object, block, and file storage solve different application requirements.
• Persistent storage survives compute restart or deletion, while ephemeral storage may not.
• IOPS, throughput, and latency are the primary storage-performance measurements.
• Durability measures data survival, while availability measures accessibility.
• Replication improves availability and resilience but does not replace backup.
• RPO and RTO define data-loss and recovery-time expectations.
• Storage tiers reduce cost by matching price and retrieval behavior to access frequency.
• Lifecycle policies automate storage transitions and deletion.
• Versioning and immutable storage improve protection against accidental or malicious changes.
• Encryption, least-privilege access, network controls, and logging are essential storage-security measures.
• Backup restoration must be tested regularly.
• Storage architecture should consider capacity, operations, retrieval, replication, and transfer cost.
• The correct storage choice depends on access pattern, performance, sharing, consistency, resilience, security, and cost.

---