Summary:
- AWS Storage Gateway provides persistent hybrid cloud access with local caching, while DataSync excels at high-speed bulk data transfers and migrations.
- How to choose between AWS Storage Gateway and DataSync based on whether you need persistent hybrid access or one-time data migration
- Why cache sizing and throughput architecture differ fundamentally between appliance-based gateways and agent-based transfer services
- When to combine both services for disaster recovery workflows that require ongoing synchronization with local file access
- What changed with the mandatory Amazon Linux 2023 migration deadline of June 30, 2026 for all Storage Gateway appliances
Choosing between AWS Storage Gateway and AWS DataSync determines whether your hybrid storage architecture will succeed or collapse under operational complexity. Both services move data between on-premises environments and AWS, yet they solve fundamentally different problems. Storage Gateway provides persistent, protocol-native access to cloud storage through local appliances. DataSync delivers high-speed, scheduled data transfers without maintaining local storage infrastructure.
The wrong choice leads to either unnecessary infrastructure costs or severe latency for daily operations. This guide breaks down the architectural differences, performance characteristics, and cost models so you can make the right decision for your workload.
What is AWS Storage Gateway
AWS Storage Gateway bridges on-premises applications to AWS cloud storage without requiring application rewrites. The service deploys as a virtual machine or hardware appliance in your data center, presenting standard storage protocols that applications already understand. Your backup software, file servers, and enterprise applications connect to Storage Gateway using NFS, SMB, or iSCSI. The gateway handles all data movement to Amazon S3, Amazon FSx, or Amazon EBS behind the scenes.
Three gateway types address distinct storage patterns. S3 File Gateway exposes Amazon S3 buckets as NFS or SMB file shares, enabling file-based workloads to store objects directly in S3 with automatic tiering across storage classes. Volume Gateway provides block storage volumes via iSCSI, storing data in S3 with optional local caching or full local storage with cloud backup. Tape Gateway emulates a virtual tape library for backup applications, archiving data to S3 Glacier storage classes. Each type maintains a local cache to deliver low-latency access for frequently used data while keeping the authoritative copy in AWS.
The service fits organizations that need continuous, transparent access to cloud storage from existing applications. Branch offices accessing centralized file shares, backup systems writing to virtual tape libraries, and applications requiring block storage with cloud durability all benefit from Storage Gateway’s protocol translation layer. The gateway handles encryption, compression, and bandwidth optimization automatically, making cloud storage appear as local infrastructure to consuming applications.
What is AWS DataSync
AWS DataSync accelerates data movement between storage systems without maintaining persistent local infrastructure. The service executes transfer tasks that copy data from source to destination, verify integrity, and complete. Unlike Storage Gateway’s always-on model, DataSync operates as a scheduled or on-demand transfer engine optimized for throughput rather than continuous access.
DataSync supports transfers between on-premises storage (NFS, SMB, HDFS, object storage), other cloud providers (Azure Blob, Google Cloud Storage), and AWS services including Amazon S3, Amazon EFS, and all Amazon FSx variants. The service deploys lightweight agents for on-premises transfers while supporting agentless operation for cloud-to-cloud movements. Each transfer task can run in Basic mode with sequential processing or Enhanced mode with parallel operations for higher performance.
Migration projects, disaster recovery replication, and data distribution workflows represent DataSync’s primary use cases. Organizations moving petabytes from data centers to AWS, replicating datasets across regions for business continuity, or synchronizing on-premises file servers with cloud storage benefit from DataSync’s optimized transfer protocol. The service handles incremental transfers automatically, copying only changed data on subsequent executions to minimize transfer time and cost.
Core architectural differences
The fundamental distinction lies in operational model, not capability. Storage Gateway maintains state through local cache and continuous synchronization. DataSync executes stateless transfer operations that complete and terminate. This difference cascades into every architectural decision from hardware requirements to cost structure.
Deployment and infrastructure requirements
Storage Gateway requires dedicated infrastructure that runs continuously. Virtual machine deployments need minimum 4 vCPUs, 16 GB RAM, and 80 GB root disk space, plus additional cache storage sized to your working set. The Storage Gateway Hardware Appliance provides a pre-configured 1U server with 5 TB or 12 TB SSD cache for environments lacking virtualization infrastructure. Cache disk sizing directly impacts performance. AWS recommends cache sized to 1.1 times your upload buffer for optimal throughput.
DataSync agents have lighter requirements but still need dedicated resources during transfers. Basic mode agents require 4 vCPUs, 32 GB RAM, and 80 GB disk space. Enhanced mode agents need 8 vCPUs with the same memory and storage. Amazon EC2 deployments should use m5.2xlarge instances minimum for Basic mode or m6a.2xlarge for Enhanced mode. Unlike Storage Gateway, DataSync agents can be stopped between transfer tasks, reducing operational overhead for infrequent migrations.
Protocol support and access patterns
Storage Gateway translates between on-premises protocols and cloud storage APIs. Applications connect using familiar interfaces.
- S3 File Gateway: NFSv3, NFSv4.1, SMBv2, SMBv3 for file-based access to S3 objects
- Volume Gateway: iSCSI for block storage with cached or stored volume modes
- Tape Gateway: iSCSI-based virtual tape library compatible with major backup applications
DataSync connects directly to storage systems without protocol translation. The service reads from source locations using native protocols (NFS, SMB, HDFS, S3 API, Azure Blob API) and writes to destinations using their native interfaces. This direct approach eliminates the latency of protocol conversion but requires applications to access destination storage directly after transfer completion.
Performance benchmarks and throughput analysis
Performance characteristics diverge based on access pattern and cache behavior. Storage Gateway throughput depends heavily on cache hit rates, while DataSync performance scales with network bandwidth and task parallelization.
Storage Gateway File Gateway performance
File Gateway throughput varies dramatically between cache hits and cache misses. With a properly sized cache (512 GiB, io1 with 1,500 IOPS) and 16 vCPUs, Linux clients achieve 590 MiB/sec (4.9 Gbps) for cache hit reads using NFSv3 with 8 threads. Cache miss reads drop to 335 MiB/sec (2.8 Gbps) as data must be retrieved from S3. Write throughput reaches 160 MiB/sec (1.3 Gbps) under the same configuration.
Scaling to high-performance configurations with 32 vCPUs, 244 GB RAM, and 4×2 TB NVMe cache disks pushes write throughput to 585 MiB/sec (4.9 Gbps) with 8 threads. The Hardware Appliance delivers 385 MiB/sec write throughput with NFSv3 on Linux clients. Windows clients using SMB see lower cache miss performance, dropping to 20 MiB/sec (0.2 Gbps) for single-threaded operations due to protocol overhead.
DataSync transfer performance
DataSync maximum throughput reaches 10 Gbps per task for agent-based transfers and 5 Gbps for agentless cloud-to-cloud transfers. Enhanced mode achieves higher effective throughput than Basic mode by parallelizing all transfer phases. A 50 TB migration using Enhanced mode completes faster than Basic mode because listing, preparation, transfer, and verification happen concurrently rather than sequentially.
The following table compares key performance characteristics.
| Metric | S3 File Gateway (cache hit) | S3 File Gateway (cache miss) | DataSync Enhanced mode | DataSync Basic mode |
|---|---|---|---|---|
| Maximum throughput | 590 MiB/sec (4.9 Gbps) | 335 MiB/sec (2.8 Gbps) | 10 Gbps (agent), 5 Gbps (agentless) | 10 Gbps (agent) |
| Object/file limit per operation | 20 million cached metadata | Limited by S3 | Virtually unlimited | 50 million (on-prem), 25 million (AWS) |
| Latency profile | Sub-millisecond (cached) | S3 retrieval latency | Batch transfer latency | Sequential processing latency |
| Best for | Repeated access to working set | Infrequent cold data access | Large migrations, unlimited scale | Smaller datasets, all location types |
Cost comparison and pricing models
Pricing structures reflect the operational differences between persistent infrastructure and task-based execution. Storage Gateway charges for gateway usage, storage consumed, and data transfer. DataSync charges per gigabyte transferred plus per-execution fees for Enhanced mode tasks.
Storage Gateway cost components
S3 File Gateway pricing includes gateway usage fees up to $125 per gateway per month, with the first 100 GB written to AWS free per account. Additional costs include the following.
- Amazon S3 storage costs based on storage class (Standard, Intelligent-Tiering, Glacier)
- S3 request charges for PUT, GET, LIST, and HEAD operations
- Data transfer out charges when retrieving data from AWS
- CloudWatch costs for monitoring and health notifications
Volume Gateway adds snapshot storage costs in Amazon S3 and potential EBS charges when restoring volumes. Tape Gateway includes virtual tape library storage, virtual tape shelf (archive) storage, and retrieval fees from Glacier storage classes. The cache infrastructure requires local storage investment, whether through virtual machine disk allocation or hardware appliance purchase.
DataSync cost components
DataSync pricing follows a simpler model. Basic mode charges $0.0125 per GB transferred. Enhanced mode charges $0.015 per GB plus $0.55 per task execution. Additional costs include S3 request charges, data transfer between regions, and Secrets Manager fees for custom credentials.
Consider a 50 TB migration to S3 Standard.
- DataSync Basic mode: (50 TB × 1024 GB × $0.0125) + S3 PUT requests ≈ $656
- DataSync Enhanced mode: (50 TB × 1024 GB × $0.015) + $0.55 + S3 requests ≈ $798
- Storage Gateway (ongoing): Up to $125/month gateway fee + S3 storage + transfer costs
For one-time migrations, DataSync costs less than maintaining Storage Gateway infrastructure. For ongoing hybrid access, Storage Gateway’s fixed monthly fee becomes economical compared to repeated DataSync transfers of the same data.
Decision framework for service selection
Selecting between Storage Gateway and DataSync requires evaluating access patterns, migration timelines, and operational requirements. Neither service universally outperforms the other. The right choice depends on how your applications interact with data after transfer.
When to choose Storage Gateway
Storage Gateway fits scenarios requiring persistent, low-latency access to cloud-stored data from on-premises applications. Select Storage Gateway when:
- Applications require NFS, SMB, or iSCSI protocol access to cloud storage
- Users need transparent file share access without knowing data resides in S3
- Backup software requires virtual tape library integration
- Block storage workloads need cloud-backed volumes with local caching
- Working set data must remain locally cached for sub-millisecond access
When to choose DataSync
DataSync excels at moving data efficiently without maintaining persistent infrastructure. Select DataSync when:
- Migrating data centers to AWS with defined project timelines
- Replicating datasets between AWS regions for disaster recovery
- Synchronizing on-premises file servers with cloud storage on a schedule
- Transferring data from other cloud providers to AWS
- Applications will access destination storage directly after transfer
When to combine both services
Complex architectures benefit from using both services together. DataSync handles initial bulk migration while Storage Gateway provides ongoing hybrid access. Disaster recovery workflows use DataSync for cross-region replication and Storage Gateway for local failover access. Data lake architectures might use DataSync to ingest data from multiple sources while Storage Gateway provides file-based access for analytics tools.
Amazon Linux 2023 migration requirements
AWS is transitioning all Storage Gateway appliances from Amazon Linux 2 to Amazon Linux 2023. This mandatory migration affects S3 File Gateway, Tape Gateway, and Volume Gateway deployments. The transition enables new hybrid cloud storage features while maintaining security standards through the updated operating system.
Key migration timeline milestones:
- October 28, 2025: New gateway deployments default to AL2023 images
- January 5, 2026: New AL2 gateway activations blocked
- June 30, 2026: Software updates and support end for AL2-based gateways
Identify affected gateways through the AWS Console, API, or AWS Health Dashboard. Migration guides exist for each gateway type, with data remaining durably stored in AWS during the process. Downtime varies by gateway type and configuration. The Storage Gateway migration documentation provides step-by-step procedures for each gateway variant.
Conclusion
AWS Storage Gateway and DataSync serve complementary roles in hybrid cloud architectures. Storage Gateway provides persistent, protocol-native access to cloud storage through local appliances with caching, making it ideal for applications requiring continuous file, block, or tape access. DataSync delivers high-throughput data transfers for migrations, replication, and scheduled synchronization without maintaining always-on infrastructure. The June 2026 Amazon Linux 2023 migration deadline adds urgency for Storage Gateway customers to plan their upgrade path.
Cost optimization requires matching service selection to access patterns. One-time migrations favor DataSync’s per-gigabyte pricing. Ongoing hybrid access justifies Storage Gateway’s fixed monthly costs. Complex architectures often combine both services, using DataSync for bulk data movement and Storage Gateway for operational access. As AWS continues expanding Enhanced mode capabilities and Storage Gateway performance, the decision framework remains constant. Choose based on whether your workload needs persistent local access or efficient point-to-point transfer.
