Cloud Services Cross-Reference: Storage
This document maps storage services across AWS, Azure, Oracle Cloud Infrastructure (OCI), and Google Cloud Platform (GCP) — covering object storage, block storage, file storage, archive and cold tiers, storage gateways, physical data transfer appliances, and backup services. All four providers offer broadly equivalent capabilities in each category, but pricing models, performance tier architectures, and hybrid-connectivity approaches differ materially. Use this reference when evaluating storage strategy across clouds or planning a migration.
1. Object Storage
Object storage is the foundational unstructured data service across all four clouds, used for backups, data lakes, static web content, log archives, and ML training data. Each provider offers 11 nines (99.999999999%) annual data durability, regional and geo-redundant replication, lifecycle policies, versioning, and event-driven notifications.
AWS — Amazon S3 (Simple Storage Service) S3 is the originating object storage service in public cloud and remains the most feature-rich. It provides a multi-tier storage class model within a single bucket, including S3 Standard, S3 Intelligent-Tiering (auto-tiering), S3 Standard-IA, S3 One Zone-IA, S3 Express One Zone (single-digit millisecond latency, single AZ), and three Glacier classes for archive. S3 supports S3 Object Lock, Replication (same-region and cross-region), Batch Operations, and server-side encryption with SSE-S3, SSE-KMS, or SSE-C.
Azure — Azure Blob Storage Azure Blob Storage serves as Azure's primary object store with four access tiers configurable at the object level: Hot, Cool, Cold, and Archive. Azure also offers Azure Data Lake Storage Gen2 (ADLS Gen2) — Blob Storage with hierarchical namespace enabled — as its optimized object store for big data analytics workloads. Azure Blob supports lifecycle management policies, versioning, immutability (WORM) policies, and integration with Azure CDN and Azure Synapse.
OCI — OCI Object Storage OCI Object Storage offers three tiers: Standard (hot), Infrequent Access (cool, 31-day minimum retention), and Archive (cold, 90-day minimum, up to 1-hour restore). The Auto-Tiering feature automatically moves objects larger than 1 MiB from Standard to Infrequent Access based on access patterns at no penalty. OCI's most significant differentiator is egress pricing: 10 TB/month free across all services in a tenancy, then a flat ~$0.0085/GB — approximately one-tenth the rate charged by AWS for comparable data volumes.
GCP — Cloud Storage Cloud Storage offers four storage classes: Standard, Nearline (30-day minimum, ~monthly access), Coldline (90-day minimum, ~quarterly access), and Archive (365-day minimum, less than annual access). All classes use the same API and bucket structure, with class determined at the object or bucket level. Autoclass automatically transitions objects between Standard, Nearline, Coldline, and Archive based on access. Cloud Storage supports Uniform Bucket-Level Access (UBLA), Object Retention Lock, and strong consistency globally.
| Function | AWS | Azure | OCI | GCP |
|---|---|---|---|---|
| Service name | Amazon S3 | Azure Blob Storage | OCI Object Storage | Cloud Storage |
| Analytics-optimized variant | No separate service | ADLS Gen2 (hierarchical namespace) | No separate service | No separate service |
| Storage tiers | Standard, Intelligent-Tiering, Standard-IA, One Zone-IA, Express One Zone, Glacier (3 classes) | Hot, Cool, Cold, Archive | Standard, Infrequent Access, Archive | Standard, Nearline, Coldline, Archive |
| Auto-tiering | S3 Intelligent-Tiering | Smart tier (preview) | Auto-Tiering | Autoclass |
| Minimum retention (infrequent) | 30 days (Standard-IA) | 30 days (Cool) | 31 days (Infrequent Access) | 30 days (Nearline) |
| Archive restore latency | Minutes (Glacier Instant) / Hours (Flexible, Deep Archive) | Up to 15 hours | Up to 1 hour | Real-time (with retrieval fee) |
| Free egress per month | 100 GB | Varies by tier/region | 10 TB | 1 TB |
| Object Lock / WORM | Yes (S3 Object Lock) | Yes (Immutability policies) | Yes (Retention Rules) | Yes (Object Retention) |
| Strong consistency | Yes | Yes | Yes | Yes |
Key differentiators:
- OCI's 10 TB/month free egress and flat per-GB rate thereafter make it significantly cheaper for data-intensive read workloads compared to AWS and Azure, which charge substantial egress fees beyond small free tiers.
- S3 Intelligent-Tiering is uniquely granular: it monitors access at the individual object level across multiple tiers including a frequent-access tier, infrequent-access tier, archive instant-access tier, archive access tier, and deep archive access tier.
- ADLS Gen2 is Azure's differentiated offering for hierarchical-namespace analytics; other providers achieve the same result via their standard object stores with analytics integrations.
- GCP's Archive class has a 365-day minimum retention period — the longest cold archive commitment of any provider.
2. Block Storage
Block storage provides persistent, high-performance disks attached to virtual machines. All providers offer SSD and HDD variants, with SSD tiers segmented by performance profile and independently provisionable IOPS and throughput. Block volumes persist independently of VM lifecycle and can be detached and reattached.
AWS — Amazon EBS (Elastic Block Store) EBS offers five volume types: gp3 (general-purpose SSD, up to 16,000 IOPS / 1,000 MB/s, IOPS and throughput provisioned independently of size), io2 Block Express (up to 256,000 IOPS / 4,000 MB/s, sub-500 microsecond latency, designed for critical databases), gp2 (legacy general-purpose SSD), st1 (throughput-optimized HDD for big data/log processing), and sc1 (cold HDD for archival). gp3 replaced gp2 as the default and is 20% cheaper at equivalent performance. EBS Multi-Attach (io2) allows up to 16 EC2 instances in the same AZ to concurrently access a single volume.
Azure — Azure Managed Disks Azure Managed Disks has five types: Ultra Disk (up to 400,000 IOPS / 10,000 MB/s, sub-millisecond latency, independently provisionable performance), Premium SSD v2 (up to 80,000 IOPS / 1,200 MB/s, 3,000 IOPS baseline free), Premium SSD v1 (up to 20,000 IOPS / 900 MB/s), Standard SSD (up to 6,000 IOPS / 750 MB/s), and Standard HDD (up to 2,000 IOPS / 500 MB/s). Azure Elastic SAN provides a cloud-native Storage Area Network service that pools capacity from multiple disk types behind a single SAN interface.
OCI — OCI Block Volumes OCI Block Volumes uses a Volume Performance Units (VPUs) model where performance is provisioned per GB, spanning four tiers: Lower Cost (0 VPU, throughput-oriented, no SLA), Balanced (10 VPU, default, balanced SSD), Higher Performance (20 VPU), and Ultra High Performance (30–120 VPU, up to 300,000 IOPS / 2,680 MB/s per volume). The elastic performance model allows IOPS and throughput to scale independently of volume size across these tiers. OCI Block Volumes supports up to 1 PB per compute instance (aggregate across attached volumes).
GCP — Persistent Disk and Hyperdisk GCP has two block storage families. Persistent Disk (legacy) includes pd-ssd, pd-balanced, pd-standard (HDD), and pd-extreme. Hyperdisk (current generation) has four types: Hyperdisk Balanced (up to 160,000 IOPS / 2,400 MiB/s, sub-millisecond latency), Hyperdisk Extreme (up to 350,000 IOPS, highest IOPS offering), Hyperdisk Throughput (optimized for Hadoop/Kafka cost-effective throughput), and Hyperdisk ML (multi-reader shared volumes for ML model serving, up to 1.6M IOPS at 100 GiB/s throughput on a single volume shared across VMs).
| Function | AWS EBS | Azure Managed Disks | OCI Block Volumes | GCP Hyperdisk |
|---|---|---|---|---|
| General-purpose SSD | gp3 | Premium SSD v2 | Balanced (10 VPU) | Hyperdisk Balanced |
| High-IOPS SSD | io2 Block Express | Ultra Disk | Ultra High Performance (30–120 VPU) | Hyperdisk Extreme |
| Max IOPS per volume | 256,000 (io2 BE) | 400,000 (Ultra) | 300,000 (UHP) | 350,000 (Extreme) |
| Max throughput per volume | 4,000 MB/s | 10,000 MB/s | 2,680 MB/s | varies by type |
| Cold/archival HDD | sc1 | Standard HDD | Lower Cost (0 VPU) | pd-standard |
| Multi-attach | Yes (io2, up to 16 instances) | Yes (Ultra, Premium SSD v2) | Yes | Yes (Hyperdisk ML multi-reader) |
| Performance model | Provisioned IOPS + throughput separate from size (gp3, io2) | Provisioned IOPS + throughput separate from size (Ultra, Premium SSD v2) | VPU per GB across tiers | Provisioned IOPS + throughput separate from size |
| SAN offering | No | Azure Elastic SAN | No | No |
| Container-native | Yes (EBS CSI) | Yes (AKS CSI) | Yes (OKE CSI) | Yes (GKE CSI, Hyperdisk for GKE) |
Key differentiators:
- GCP Hyperdisk ML is unique: a single volume can be mounted read-only by many VMs simultaneously with up to 1.6M IOPS, designed specifically for ML inference serving where many GPU VMs need low-latency access to the same model weights.
- Azure Ultra Disk provides the highest max throughput (10,000 MB/s) of any block storage product across the four providers.
- OCI's VPU model lets operators dial in performance granularly per GB independent of absolute volume size, which is useful for right-sizing without over-provisioning capacity.
- AWS EBS Multi-Attach (io2) is limited to 16 instances; this is a shared-nothing model requiring cluster-aware file systems or applications.
3. File Storage
Managed file storage exposes NFS or SMB shares to multiple compute instances simultaneously without requiring customers to manage file server infrastructure. Use cases include home directories, CMS platforms, ERP shared data, HPC scratch space, and lift-and-shift of on-premises NAS workloads.
AWS — Amazon EFS and Amazon FSx AWS offers two distinct file storage services. Amazon EFS is a serverless, elastic NFS file system that scales automatically from near-zero to petabytes, charging only for data stored. EFS supports Standard and Infrequent Access storage classes with automatic lifecycle management. Amazon FSx provides four fully managed file systems targeting specific protocols and workloads: FSx for Windows File Server (SMB, Active Directory integration), FSx for NetApp ONTAP (NFS, SMB, iSCSI, multi-protocol), FSx for OpenZFS (NFS, Linux/Windows/macOS), and FSx for Lustre (high-performance parallel file system for HPC and ML, single-AZ only).
Azure — Azure Files and Azure NetApp Files Azure Files provides fully managed SMB and NFS file shares in four performance tiers: Transaction Optimized, Hot, Cool, and Premium (SSD-backed). Azure NetApp Files (ANF) is a joint offering with NetApp, delivering enterprise-grade NFS (v3/v4.1) and SMB workloads with sub-millisecond latency and up to Ultra performance tier (4,500 MiB/s per volume). Azure File Sync enables caching of Azure Files on-premises Windows Servers, functioning as a cloud-tiered NAS extension. Azure Container Storage provides persistent volume management for stateful AKS workloads backed by Azure Disks, Elastic SAN, or Azure Files.
OCI — OCI File Storage and File Storage with Lustre OCI File Storage is a fully managed, elastic NFS v3 file system scaling automatically to 8 exabytes. It supports snapshots (up to 10,000 per file system), cross-region replication, and mount targets within a VCN for private access. OCI File Storage with Lustre is a separate, fully managed parallel file system service delivering petabyte capacity and terabytes-per-second throughput for AI and HPC workloads, without requiring Lustre cluster management expertise.
GCP — Filestore and NetApp Volumes Google Cloud Filestore provides managed NFS file shares in three service tiers: Basic HDD, Basic SSD, and Enterprise (high availability, multi-zone, snapshots, backup). Google Cloud NetApp Volumes (CVS) is a fully managed file service based on NetApp ONTAP, offering NFS (v3/v4.1) and SMB with snapshots, cloning, and SnapMirror replication. Google Cloud Managed Lustre is a fully managed parallel file system optimized for AI and HPC workloads (GA in 2025).
| Function | AWS | Azure | OCI | GCP |
|---|---|---|---|---|
| Managed NFS (general) | Amazon EFS | Azure Files (NFS) | OCI File Storage | Filestore |
| Managed SMB | FSx for Windows File Server | Azure Files (SMB) | None native | None native |
| NetApp ONTAP managed | FSx for NetApp ONTAP | Azure NetApp Files | None | Google Cloud NetApp Volumes |
| High-performance parallel (Lustre) | FSx for Lustre | None | File Storage with Lustre | Managed Lustre |
| Multi-protocol (NFS + SMB + iSCSI) | FSx for NetApp ONTAP | Azure NetApp Files | None native | Google Cloud NetApp Volumes |
| Auto-scaling | Yes (EFS elastic) | No (fixed capacity) | Yes (elastic, up to 8 EB) | Yes (Enterprise tier) |
| On-premises caching/sync | None native | Azure File Sync | None native | None native |
| Serverless/pay-per-GB | Yes (EFS) | No | Yes | No |
Key differentiators:
- AWS FSx provides the broadest choice of file system personalities (Windows, ONTAP, OpenZFS, Lustre) under a single managed service umbrella, which is uniquely suited for enterprises running mixed OS workloads.
- Azure File Sync has no direct equivalent at other providers: it turns any Windows Server into a tiered cache of an Azure Files share, enabling hybrid on-premises file server consolidation without moving all data to cloud.
- OCI File Storage's 8 EB elastic ceiling and 10,000 snapshots per file system are standout limits for large-scale enterprise deployments.
- All four providers now offer a managed Lustre option (FSx for Lustre, OCI File Storage with Lustre, GCP Managed Lustre), closing a gap that previously required self-managed HPC clusters.
4. Archive and Cold Storage Tiers
All four providers offer deeply reduced-cost storage tiers for data that is accessed rarely or not at all. The tradeoff is higher retrieval costs and, in some cases, mandatory minimum retention periods and delayed access.
AWS — S3 Glacier classes AWS has three Glacier tiers within S3: S3 Glacier Instant Retrieval (millisecond access, 90-day minimum retention, same API as standard S3 objects), S3 Glacier Flexible Retrieval (minutes to 12-hour retrieval with three retrieval speeds: Expedited, Standard, Bulk; 90-day minimum), and S3 Glacier Deep Archive (lowest-cost storage in AWS at ~$1/TB-month; 12–48 hour retrieval; 180-day minimum). Virtual Tape Library backups stored via Tape Gateway are archived to Glacier Flexible Retrieval or Deep Archive.
Azure — Blob Archive tier Azure Blob Storage Archive tier is offline storage (data is not immediately accessible) requiring rehydration to Hot, Cool, or Cold before reading; rehydration takes up to 15 hours at standard priority or hours at high priority. Archive has the lowest storage cost in Azure and a 180-day minimum retention period. Azure also supports Blob Lifecycle Management policies to automatically transition blobs through Hot → Cool → Cold → Archive based on age or last-access time.
OCI — OCI Archive Storage OCI Archive Storage is the primary service for cold data — it is distinct from Object Storage and uses the Archive tier bucket type. Objects in Archive require explicit restore before access, with restore completing within 1 hour. The minimum retention period is 90 days. OCI Archive Storage is priced as the lowest per-GB cost in OCI's storage portfolio. Within Object Storage buckets, individual objects can also be set to the Archive tier using lifecycle policies.
GCP — Cloud Storage Archive class Cloud Storage Archive has the lowest storage cost on GCP and a 365-day minimum retention period — the longest of any provider — making it suitable only for true long-term retention workloads. Retrieval is real-time (sub-millisecond API access) but incurs retrieval charges, unlike the offline/restore model used by AWS Glacier Flexible and Azure Archive. The 365-day minimum means early deletion results in prorated charges.
| Function | AWS | Azure | OCI | GCP |
|---|---|---|---|---|
| Archive service | S3 Glacier (3 classes) | Blob Storage Archive tier | OCI Archive Storage | Cloud Storage Archive class |
| Lowest-cost tier approx. price | ~$0.00099/GB-month (Deep Archive) | ~$0.001/GB-month | Lower than Infrequent Access (~$0.0026/GB-month) | ~$0.0012/GB-month |
| Minimum retention | 90 days (Instant/Flexible), 180 days (Deep Archive) | 180 days | 90 days | 365 days |
| Restore / access model | Instant (Instant Retrieval) or delayed restore (hours) | Offline — must rehydrate (up to 15 hours) | Offline — restore completes within 1 hour | Real-time access with retrieval fee |
| Retrieval speeds | Expedited (~1–5 min), Standard (~3–5 hr), Bulk (~5–12 hr) for Flexible | Standard (~15 hr), High Priority (< 1 hr) | Single tier, ≤1 hour | Real-time |
| Tape emulation | Yes (Tape Gateway → Glacier) | No | No | No |
| Lifecycle policy automation | Yes (S3 Lifecycle) | Yes (Blob Lifecycle Management) | Yes (Object Lifecycle Management) | Yes (Lifecycle policies / Autoclass) |
Key differentiators:
- S3 Glacier Instant Retrieval is the only true archive tier with millisecond access — it behaves like a standard object store but is priced as archive. This eliminates the restore-wait problem for data that is infrequent but occasionally time-sensitive.
- GCP Archive's real-time access model (no restore step) is convenient but carries the longest minimum retention commitment (365 days); organizations with uncertain access patterns should evaluate this carefully.
- OCI Archive Storage's 1-hour restore SLA is competitive with Azure's and is faster for Bulk data than AWS Glacier Flexible Retrieval Bulk tier.
5. Storage Gateways
Storage gateways provide on-premises applications with transparent access to cloud storage using familiar protocols (NFS, SMB, iSCSI, virtual tape). They are deployed as VM appliances or hardware devices at the customer site.
AWS — AWS Storage Gateway AWS Storage Gateway offers four gateway types: S3 File Gateway (NFS/SMB → S3 object storage), FSx File Gateway (SMB → FSx for Windows File Server, local cache), Volume Gateway (iSCSI block storage backed by S3; Cached mode keeps frequently accessed data on-premises, Stored mode keeps all data locally and async replicates to S3), and Tape Gateway (iSCSI Virtual Tape Library backed by S3 and Glacier). Gateways deploy as VM appliances on VMware ESXi, KVM, Hyper-V, or as an EC2 instance, or as a hardware appliance.
Azure — Azure Stack Edge and Azure Data Box Gateway Azure's storage gateway story centers on two products. Azure Data Box Gateway is a software-only virtual appliance that exposes NFS and SMB shares locally while transparently syncing data to Azure Blob Storage or Azure Files — it is Azure's direct equivalent to AWS S3 File Gateway. Azure Stack Edge is a physical Microsoft-managed appliance that adds local compute (Kubernetes, Azure IoT Edge, GPU options) alongside the gateway function, enabling edge inference and data processing before transfer. Azure File Sync (see File Storage section) provides hybrid caching for Azure Files specifically.
OCI — OCI Storage Gateway OCI Storage Gateway is a Docker-based virtual appliance that exposes a local NFS v4 mount point and transparently writes data to OCI Object Storage buckets. It is commonly deployed for RMAN Oracle database backups to cloud, eliminating tape infrastructure. The gateway supports streaming and bulk transfer modes and includes local caching for recently accessed objects.
GCP — No dedicated storage gateway product GCP does not offer a standalone on-premises storage gateway service equivalent to AWS Storage Gateway or Azure Stack Edge. Organizations connecting on-premises workloads to Cloud Storage typically use Storage Transfer Service (for bulk or scheduled transfers), the gsutil/gcloud CLI tools, or partner solutions (e.g., NetApp StorageGRID, Panzura). For hybrid file access, Google Cloud NetApp Volumes with SnapMirror from on-premises ONTAP is the closest equivalent for enterprise NAS migrations.
| Function | AWS | Azure | OCI | GCP |
|---|---|---|---|---|
| File gateway (NFS/SMB → object) | S3 File Gateway | Azure Data Box Gateway | OCI Storage Gateway (NFS → Object Storage) | No native service |
| Tape gateway (VTL → cloud archive) | Tape Gateway (→ S3/Glacier) | None | None | None |
| Block gateway (iSCSI → cloud) | Volume Gateway (→ S3) | None | None | None |
| Edge compute + gateway | AWS Snowball Edge (some overlap) | Azure Stack Edge | None | None |
| Deployment model | VM appliance / hardware appliance | Virtual appliance / physical appliance | Docker container | N/A |
| Primary use case | Hybrid storage; tape modernization; backup | Hybrid file sync; edge compute/AI | Oracle DB backup (RMAN); object store access | Bulk transfer via Storage Transfer Service |
Key differentiators:
- AWS Storage Gateway has the most complete feature set: four distinct gateway types covering file, block, and tape protocols, with flexible deployment options.
- OCI Storage Gateway's Docker-based deployment is lightweight and easy to integrate into existing Linux environments, particularly for Oracle RMAN backup targets.
- Azure Stack Edge uniquely combines a storage gateway with a managed edge compute platform (Kubernetes, GPU support) for environments needing local AI inference before data transfer.
- GCP's absence of a native gateway product is a gap; the recommendation is Storage Transfer Service for scheduled/bulk online transfers, or partner appliances for persistent on-premises NFS/SMB access to Cloud Storage.
6. Data Transfer and Migration Appliances
Physical transfer appliances address the bandwidth gap when migrating petabyte-scale datasets to cloud over the public internet is impractical due to time, cost, or security constraints.
AWS — AWS Snow Family AWS Snow Family includes Snowcone (8 TB HDD or 14 TB SSD, portable, battery-operated, smallest form factor), Snowball Edge Storage Optimized (210 TB NVMe, up to 1.5 GB/s transfer rate, supports 2 PB/month migration velocity), Snowball Edge Compute Optimized (28 TB NVMe + 104 vCPUs / 416 GB RAM for edge compute with local S3-compatible storage), and Snowmobile (100 PB per truck, for datacenter-scale migrations). AWS DataSync is the online complement: an agent-based service for accelerated network transfers from on-premises or other clouds to S3/EFS/FSx. Note: As of November 2025, Snowball Edge devices are only available to existing customers; new customers are directed to AWS DataSync or AWS Data Transfer Terminal.
Azure — Azure Data Box Family Azure Data Box family includes Data Box Disk (up to 35 TB per order via SSDs), Data Box (100 TB per device, NAS-like interface), Data Box Heavy (1 PB per device, rack-mounted, for largest migrations), and Data Box Gateway (virtual appliance, no physical device, continuous online sync). Azure Import/Export Service is the legacy self-managed equivalent where customers ship their own drives. Azure Migrate is the broader migration platform that orchestrates assessment and migration of VMs, databases, and web apps, with built-in integration to Data Box for large data sets.
OCI — OCI Data Transfer Service and Roving Edge Infrastructure OCI Data Transfer Service supports both disk-based and appliance-based offline transfer. Customers can ship their own hard disks to an Oracle transfer site, or request Oracle-supplied Data Transfer Appliances for petabyte-scale migrations; Oracle uploads the data to the designated Object Storage bucket. For edge environments with limited or no WAN connectivity, the Roving Edge Device (RED) is a ruggedized, portable server that runs OCI services locally (Compute, Object Storage, Block Volume) and synchronizes to OCI over available bandwidth; the Roving Edge Ultra (REU) is a backpack-form-factor single-device variant for individual operators in austere environments. Online data transfer to OCI is supported via FastConnect (dedicated private connectivity) or the internet.
GCP — Google Cloud Transfer Appliance and Storage Transfer Service Google Cloud Transfer Appliance is available in two sizes: 100 TB and 480 TB (the older TA7 model has been deprecated). All appliances use AES-256 encryption and NIST 800-88 compliant erasure after upload. Data is shipped to Google upload facilities (EU customers ship to Belgium). Storage Transfer Service is the online/scheduled transfer service supporting transfers from AWS S3, Azure Blob Storage, HTTP/HTTPS sources, on-premises POSIX file systems, and other Cloud Storage buckets. It handles petabyte-scale transfers with bandwidth scheduling, filtering, and incremental sync.
| Function | AWS | Azure | OCI | GCP |
|---|---|---|---|---|
| Small-form appliance (< 20 TB) | Snowcone (8/14 TB) | Data Box Disk (35 TB via SSDs) | Disk-based transfer (customer-supplied disks) | No equivalent (TA7 deprecated) |
| Mid-range appliance (~100 TB) | Snowball Edge (210 TB) | Data Box (100 TB) | Oracle Data Transfer Appliance | Transfer Appliance (100 TB) |
| Petabyte appliance | Snowmobile (100 PB/truck) | Data Box Heavy (1 PB) | Oracle Data Transfer Appliance (petabyte) | Transfer Appliance (480 TB) |
| Edge compute on appliance | Yes (Snowball Edge Compute Optimized) | Yes (Azure Stack Edge) | Yes (Roving Edge Device) | No |
| Online/scheduled transfer service | AWS DataSync | Azure Data Box Gateway / Azure Migrate | FastConnect / internet | Storage Transfer Service |
| Ruggedized/field-deployable | Snowcone (portable) | None | Roving Edge Ultra (backpack) | None |
Key differentiators:
- AWS Snowmobile (100 PB per truck) remains unique for true datacenter-level migrations with no equivalent at other providers at comparable scale.
- OCI Roving Edge Device and Roving Edge Ultra serve a distinct use case: persistent edge cloud deployments in disconnected environments, not just one-time data transfer. They run actual OCI services locally, making them more than a migration appliance.
- GCP Storage Transfer Service is the most capable online transfer service for multi-cloud ingestion, natively supporting direct migration from AWS S3 and Azure Blob without deploying agents.
- Azure Data Box Heavy (1 PB) has a larger capacity than any single Snowball Edge device, though Snowball Edge appliances can be clustered.
7. Backup Services
Centralized backup services automate data protection policies across multiple resource types, enforce retention schedules, manage recovery points, and support compliance requirements. All four providers offer cross-region backup replication.
AWS — AWS Backup AWS Backup provides a centralized policy engine for backing up EC2 instances, EBS volumes, RDS databases (all engines), Aurora, DynamoDB, EFS, FSx (all flavors), S3, Storage Gateway volumes, VMware Cloud on AWS VMs, and more. Backup plans define backup frequency, retention, and cross-region/cross-account copy rules. Backup vaults are immutable by default; logically air-gapped vaults store backups in an AWS-owned account using AWS-owned keys or customer-managed KMS keys, providing isolation from the source account. AWS Backup Audit Manager provides compliance reporting against frameworks such as PCI, HIPAA, and custom controls. Item-level recovery (GA 2025) allows restoring individual files or S3 objects from a backup without restoring the full resource.
Azure — Azure Backup Azure Backup uses two vault types: Recovery Services Vault (supports Azure VMs, SQL Server in VMs, SAP HANA, Azure Files, on-premises workloads via MARS agent, and Azure Site Recovery integration) and Backup Vault (newer vault type for Azure Disks, Blobs, AKS, PostgreSQL). Azure Site Recovery (ASR) is the disaster recovery service, providing VM replication and orchestrated failover between Azure regions or from on-premises to Azure. Azure Backup Center provides a unified management view across all vaults and subscriptions. Soft delete retains backup data for 14 days after deletion at no additional cost.
OCI — OCI Backup (per-service) OCI does not offer a single unified backup service equivalent to AWS Backup. Instead, each storage service manages its own backup capabilities. Block Volume Backup supports incremental and full backups to Object Storage with predefined Gold, Silver, and Bronze policies (configurable frequency from daily to monthly with retention periods), plus on-demand manual backups. File Storage supports up to 10,000 snapshots per file system and data can be replicated using rsync or rclone for cross-region DR. Object Storage data protection relies on Cross-Region Replication and lifecycle management. OCI has no equivalent to AWS Backup's unified cross-service policy engine.
GCP — Google Cloud Backup and DR Service Google Cloud Backup and DR Service (formerly Actifio-based) provides application-consistent, centralized backup for Compute Engine VMs, VMware VMs on GCP (GCVE), file systems, and databases (Oracle, SQL Server, MySQL, SAP HANA, PostgreSQL, MongoDB, and others). Backup vaults enforce immutable retention policies; backups are indelible for the specified enforced retention period. Multi-region backup vaults (GA 2025) replicate backup data across geographic regions within a single vault resource. Backup for GKE provides native Kubernetes workload backup including application state and persistent volumes. Cloud SQL Enhanced Backups (GA 2025) integrates Backup and DR Service directly into Cloud SQL.
| Function | AWS | Azure | OCI | GCP |
|---|---|---|---|---|
| Unified backup service | AWS Backup | Azure Backup | No (per-service) | Backup and DR Service |
| VM backup | Yes (EC2) | Yes (Azure VMs) | Yes (Block Volume/Boot Volume backup) | Yes (Compute Engine VMs) |
| Database backup | Yes (RDS, Aurora, DynamoDB, DocumentDB) | Yes (SQL Server, SAP HANA, PostgreSQL) | Yes (per-service; RDBCS via Block Volume backup) | Yes (Oracle, SQL Server, MySQL, SAP HANA, etc.) |
| Kubernetes backup | No native (use EBS snapshots) | Yes (AKS backup via Backup Vault) | No native | Yes (Backup for GKE) |
| Cross-region replication | Yes | Yes (GRS / Cross-Region Restore) | Yes (Block Volume cross-region replication) | Yes (multi-region vaults) |
| Cross-account isolation | Yes (logically air-gapped vault) | Yes (separate subscriptions) | No dedicated feature | Yes (separate projects) |
| Immutability / air-gapped vault | Yes (logically air-gapped vault) | Yes (immutable vault) | No dedicated feature | Yes (enforced retention vaults) |
| Compliance reporting | Yes (Backup Audit Manager) | Yes (Azure Policy integration) | Limited | Limited |
| DR/failover orchestration | AWS Elastic Disaster Recovery | Azure Site Recovery | No native service | No native service (manual or partner) |
Key differentiators:
- AWS Backup and GCP Backup and DR Service both offer unified, multi-service backup policy engines. OCI's per-service model requires operators to manage backup policies separately for Block Volumes, File Storage, and databases, increasing operational overhead in large environments.
- Azure Site Recovery is a mature, production-grade DR orchestration platform with no direct native equivalent at AWS (AWS Elastic Disaster Recovery is the closest analog), OCI, or GCP.
- GCP's Backup for GKE is the most tightly integrated Kubernetes backup solution, with application-consistent backup of stateful workloads including persistent volumes. AWS lacks a native equivalent.
- OCI Block Volume's Gold/Silver/Bronze predefined backup policies make it easy to apply standard retention schedules without writing custom rules, but the scope is limited to block volumes only.
References
- AWS Storage Products Overview
- Amazon S3 Storage Classes
- Amazon EBS Volume Types
- Amazon FSx — Choose a File System
- AWS Storage Gateway
- AWS Snow Family
- AWS Backup
- Azure Storage Products
- Azure Blob Storage Access Tiers
- Azure Managed Disk Types
- Azure Data Box Gateway Overview
- OCI Storage Services Overview
- OCI Object Storage Tiers
- OCI Block Volume Performance
- OCI Data Transfer Service Overview
- OCI Roving Edge Infrastructure
- Google Cloud Storage Documentation
- Google Cloud Storage Classes
- Google Cloud Hyperdisk Overview
- Google Cloud Backup and DR Service
- Google Cloud Transfer Appliance