Somewhere on almost every SAA-C03 exam sits a question that reads like a logistics puzzle: a company has this much data, this much bandwidth, this deadline, and this access pattern — now pick the AWS service that gets the data where it needs to go. There are only a handful of right answers, but the wrong ones are engineered to look plausible. Choose Snowball when a scheduled DataSync task would have been faster, or reach for Storage Gateway when the scenario actually calls for a one-time bulk copy, and you lose the point.
The good news is that these questions collapse into a small decision tree once you stop memorizing service names and start reasoning about two axes: is the transfer online or offline, and is it a one-time migration or ongoing hybrid access? This guide builds that tree. We’ll cover the four services the SAA-C03 tests most — the Snow Family, AWS DataSync, AWS Storage Gateway, and AWS Transfer Family — with the bandwidth math that decides online-vs-offline, a decision table, and the scenario cues that map a paragraph to the correct answer. (Database migration with DMS is a related but separate topic; we’ll say exactly where it fits and point you elsewhere for it.)
The Two Questions Behind Every Transfer Scenario
Before you can pick a service, answer these:
- Online or offline? Can the data realistically travel over your network connection in the time available, or is it faster to physically ship it? This is a bandwidth-and-deadline calculation, not a preference.
- One-time or ongoing? Are you moving a dataset once (a migration) or do on-premises systems need continuous, low-latency access to data that lives in AWS (hybrid)?
Those two questions sort the four services cleanly:
| One-time / bulk migration | Ongoing / hybrid access | |
|---|---|---|
| Offline (ship it) | Snow Family | — |
| Online (over the network) | DataSync | Storage Gateway (block/file/tape), Transfer Family (partner uploads) |
Keep this 2×2 in your head and most exam scenarios answer themselves. The rest of the guide fills in the detail you need to break ties.
The Bandwidth Math: When Offline Beats the Wire
The single most useful skill here is estimating how long a network transfer will take. The rough formula:
transfer time (seconds) ≈ (data in bits) / (usable bandwidth in bits per second)Real links never run at 100% — assume you get 70–80% of the rated speed. A few reference points worth internalizing:
| Data set | Over 100 Mbps (~10 MB/s usable) | Over 1 Gbps (~100 MB/s usable) |
|---|---|---|
| 10 TB | ~12 days | ~1.2 days |
| 100 TB | ~4 months | ~12 days |
| 500 TB | ~1.5 years | ~2 months |
The pattern is obvious: at petabyte scale, or on a modest connection, the network is hopeless and you ship the data. AWS’s own rule of thumb is that if an online transfer would take more than about a week, you should evaluate the Snow Family. The exam loves this trigger — a scenario that mentions terabytes-to-petabytes, a slow or saturated link, or a bandwidth-constrained remote site is pointing you offline.
Conversely, if the data fits comfortably through the pipe in the available window — especially over an existing AWS Direct Connect or a decent internet link — an online service like DataSync is faster end to end because there’s no shipping round-trip.
AWS Snow Family: Offline Bulk Transfer
The Snow Family is a set of physical, ruggedized devices AWS ships to you. You load your data locally, ship the device back, and AWS imports it into Amazon S3. It’s the answer whenever the network can’t move the data in time.
| Device | Rough usable capacity | Best for |
|---|---|---|
| AWS Snowcone | ~8 TB (HDD) / ~14 TB (SSD) | Small, portable, edge or space-constrained sites; can even ship via a courier |
| Snowball Edge Storage Optimized | ~80 TB (larger 210 TB variant available) | Bulk data migration — the workhorse for tens of TB |
| Snowball Edge Compute Optimized | ~28 TB + more compute (GPU option) | Edge processing (ML inference, video analysis) before transfer |
A few properties the exam expects you to know:
- Encryption is automatic. Data is encrypted with keys managed through AWS KMS; the keys are never stored on the device, so a lost device doesn’t expose data.
- Snowball Edge is also a compute platform. The Compute Optimized devices run EC2 instances and Lambda at disconnected edge sites, so the family is about edge compute plus transfer, not transfer alone.
- Multiple devices in parallel. For very large migrations you order several devices and load them concurrently.
Accuracy note: The petabyte-scale AWS Snowmobile (the shipping-container truck) has been discontinued and is no longer part of the Snow Family. If an exam-style question offers Snowmobile for an exabyte transfer, treat it as a legacy distractor; today the answer for extreme scale is multiple Snowball Edge devices or a dedicated Direct Connect strategy.
Choose Snow when: the data is large relative to bandwidth, the deadline is tight, the site is bandwidth-constrained, or you need to pre-process data at a disconnected edge before importing it to S3.
AWS DataSync: Managed Online Transfer & Sync
AWS DataSync is a fully managed service that moves large amounts of data over the network, quickly and with integrity checking built in. Where a hand-rolled rsync or aws s3 cp script is fragile and slow, DataSync is purpose-built: AWS advertises transfers up to roughly 10× faster than open-source tooling, with automatic retries, encryption in transit, and end-to-end verification.
The mental model is an agent plus a task:
- The agent is a lightweight VM (VMware, Hyper-V, KVM, or an EC2 AMI — Snowcone ships with one preinstalled) that you deploy next to your on-premises storage to read from NFS, SMB, HDFS, or self-managed object storage. For AWS-to-AWS transfers you don’t need an agent at all.
- The task defines the source and destination, a schedule, filters, and options like bandwidth throttling.
Supported destinations cover the AWS storage that matters for SAA scenarios: Amazon S3 (any storage class), Amazon EFS, and Amazon FSx (for Windows File Server, Lustre, NetApp ONTAP, and OpenZFS).
DataSync shines for both one-time migrations and recurring transfers — for example, nightly replication of on-prem file shares into S3 for analytics, or ongoing archive to a cheaper storage class. Because you can schedule tasks and throttle bandwidth, it plays nicely with production networks and can run over Direct Connect for predictable throughput.
Choose DataSync when: you’re moving data online between on-prem NFS/SMB/HDFS and AWS (or between AWS storage services), especially when you need scheduling, filtering, throttling, or built-in integrity verification — and the network can carry the load in time.
AWS Storage Gateway: Ongoing Hybrid Access
DataSync and Snow move data; AWS Storage Gateway bridges on-premises applications to AWS storage so they keep working with local protocols while the data actually lives in AWS. It’s the hybrid answer — for workloads that need continuous access, not a one-shot copy. It comes in three modes, and knowing which is which is a classic exam discriminator:
| Gateway mode | Protocol presented on-prem | Where data lives | Typical use |
|---|---|---|---|
| S3 File Gateway | NFS / SMB file shares | Files stored as objects in Amazon S3 | Give file-based apps cloud-backed storage; ingest files into S3 |
| FSx File Gateway | SMB file shares | Amazon FSx for Windows File Server | Low-latency on-prem access to a fully managed Windows file share |
| Volume Gateway | iSCSI block volumes | Amazon S3, backed up as EBS snapshots | Block storage for on-prem servers with cloud backup |
| Tape Gateway | Virtual Tape Library (iSCSI VTL) | S3 / S3 Glacier / Glacier Deep Archive | Replace physical tape backups without changing backup software |
Volume Gateway has two sub-modes you must distinguish:
- Cached volumes — the primary copy lives in S3; a local cache holds only frequently accessed (hot) data. Best when you have lots of data but limited on-prem storage.
- Stored volumes — the primary copy stays on-premises for low-latency access; AWS holds an asynchronous backup as EBS snapshots. Best when you need the entire dataset locally but want durable cloud backup.
A reliable exam cue: cached = primary in the cloud, hot data cached locally; stored = primary local, backup in the cloud. If the scenario stresses low-latency access to the full dataset, that’s stored; if it stresses limited local capacity or cloud-backed expansion, that’s cached.
Choose Storage Gateway when: on-premises applications need ongoing access to AWS-backed storage using existing file, block, or tape workflows — a hybrid pattern, not a migration.
AWS Transfer Family: Managed SFTP, FTPS, FTP & AS2
AWS Transfer Family is a fully managed service for transferring files into and out of Amazon S3 or Amazon EFS over the protocols your partners already use: SFTP, FTPS, FTP, and AS2. The scenario signature is unmistakable — external partners, vendors, or customers upload files on a schedule using SFTP and you don’t want to run and patch your own file-transfer servers.
The value proposition is that you keep the standard protocol and endpoint your partners expect, while the data lands directly in durable, cheap S3 (or EFS) where the rest of your pipeline can pick it up. There’s no fleet of SFTP servers to maintain, and you pay per provisioned endpoint-hour plus data transferred.
Choose Transfer Family when: the requirement is inbound or outbound file exchange with third parties over a standard protocol (especially SFTP), landing in S3/EFS, with no servers to manage. It is not a bulk-migration tool — it’s a managed protocol endpoint.
Where DMS Fits (and Why It’s Out of Scope Here)
You’ll sometimes see AWS Database Migration Service (DMS) offered as an option in these questions. DMS is for migrating databases — homogeneous (e.g., Oracle → Oracle) or heterogeneous (e.g., Oracle → Amazon Aurora, paired with the Schema Conversion Tool) — with the option to keep the source running via ongoing replication. If the scenario is about files, objects, or block storage, DMS is the wrong answer; if it’s about relational or NoSQL database records, DMS is likely right and the four services above are distractors.
Database and large-scale portfolio migration have their own decision framework — the seven common migration strategies (the “7 Rs”) and the enterprise tooling around them are covered in our deep dive on AWS migration strategies for the SAP-C02 exam, which is worth a read even at the associate level for the bigger picture.
Decision Table: Which Data-Transfer Service?
| Scenario cue | Answer |
|---|---|
| Petabytes to move, network would take months | Snow Family (Snowball Edge) |
| Remote site, very limited bandwidth, ship it | Snow Family (Snowcone / Snowball Edge) |
| Need to process data at a disconnected edge before import | Snowball Edge Compute Optimized |
| Recurring, scheduled copy of on-prem NFS/SMB shares into S3 | DataSync |
| One-time online migration with integrity verification | DataSync |
| On-prem app needs NFS/SMB access, files stored in S3 | S3 File Gateway |
| Low-latency on-prem access to a managed Windows file share | FSx File Gateway |
| iSCSI block volumes, entire dataset local, cloud backup | Volume Gateway (stored) |
| iSCSI block volumes, limited local storage, primary in cloud | Volume Gateway (cached) |
| Replace physical tape backups, keep backup software | Tape Gateway |
| Partners upload files via SFTP into S3 | Transfer Family |
| Migrate an Oracle database to Aurora | DMS (out of scope here) |
Common SAA-C03 Scenarios
Scenario 1 — Petabyte data lake, tight deadline. A media company must move 900 TB of archive footage to S3 within a month; their site has a 200 Mbps link. The wire would take the better part of a year, so the answer is multiple Snowball Edge Storage Optimized devices loaded in parallel.
Scenario 2 — Nightly analytics feed. An on-prem NFS share of ~2 TB must land in S3 every night for a data pipeline, over an existing Direct Connect. That’s an online, recurring transfer with scheduling — DataSync.
Scenario 3 — Legacy app, limited local disk. An on-premises application writes to iSCSI block storage, the team is running out of local capacity, and they want the bulk of the data in the cloud with only hot blocks cached locally. That’s Volume Gateway in cached mode.
Scenario 4 — Vendor file drops. A retailer’s suppliers push daily inventory files via SFTP and the team wants those files in S3 with no servers to manage. That’s Transfer Family.
Common Mistakes to Avoid
- Reaching for Snow when the network is fine. If the data fits through the pipe in the available time, DataSync is faster overall — no shipping delay. Do the bandwidth math.
- Confusing DataSync with Storage Gateway. DataSync moves data (migration/sync); Storage Gateway provides ongoing hybrid access. “Copy once” vs. “keep accessing” is the tell.
- Mixing up cached and stored volumes. Remember: stored keeps the primary copy on-premises; cached keeps the primary in S3.
- Using Transfer Family for bulk migration. It’s a managed protocol endpoint for partner file exchange, not a petabyte mover.
- Answering DMS for file/object scenarios. DMS is databases only.
Practice Until the Decision Is Reflexive
Reading the decision table is enough to answer a clean question, but the SAA-C03 deliberately blends cues — a paragraph might combine a bandwidth constraint, a deadline, and an access pattern and expect you to weigh all three under time pressure. That fluency only comes from working realistic scenarios.
Sailor.sh’s AWS Certified Solutions Architect – Associate (SAA-C03) Mock Exam Bundle gives you exam-style transfer and hybrid-storage scenarios that mirror the real format and difficulty — including the DataSync-vs-Snow and cached-vs-stored traps covered here — with a detailed explanation on every question so a wrong answer turns into a lesson. Use the free concepts here to build the model; use timed mock exams to make the choice automatic. Round out your prep with the SAA-C03 exam guide for 2026 and a structured SAA-C03 study plan, and revisit the AWS storage services decision guide so you know exactly which destination each transfer service is feeding.
Frequently Asked Questions
When should I use Snowball instead of DataSync?
Do the bandwidth math. If moving the data online would take more than roughly a week — because the dataset is large, the deadline is tight, or the site’s connection is slow — ship it with the Snow Family. If the network can carry the load in the available window (especially over Direct Connect), DataSync is faster end to end because there’s no shipping round-trip.
What’s the difference between cached and stored Volume Gateway?
With cached volumes, the primary copy of your data lives in Amazon S3 and only frequently accessed (hot) data is cached on-premises — ideal when local storage is limited. With stored volumes, the primary copy stays on-premises for low-latency access to the full dataset, and AWS keeps an asynchronous backup as EBS snapshots. Cue: stored = primary local; cached = primary in the cloud.
Does DataSync require an agent?
Only for on-premises or self-managed sources. To read from on-prem NFS, SMB, or HDFS you deploy a lightweight DataSync agent VM next to the storage. For transfers between AWS services (for example, S3 to EFS), no agent is required.
Which Storage Gateway mode do I pick for tape backups?
Tape Gateway. It presents a Virtual Tape Library over iSCSI so your existing backup software keeps writing “tapes,” but the data is stored durably in S3, S3 Glacier, or Glacier Deep Archive — no physical tapes or tape hardware.
Is AWS Transfer Family a data-migration service?
No. Transfer Family is a fully managed endpoint for SFTP, FTPS, FTP, and AS2 that lands files in S3 or EFS — ideal for exchanging files with external partners without running your own servers. For bulk one-time migration use DataSync (online) or Snow (offline).
Where does DMS fit among these services?
AWS Database Migration Service migrates databases, not files or objects — for example, moving an on-prem Oracle database to Amazon Aurora, optionally with ongoing replication so the source stays live during cutover. If a scenario is about database records rather than files, DMS is the likely answer and the transfer services above are distractors.
Conclusion
AWS gives you a service for every shape of data movement, and the SAA-C03 rewards you for matching the shape to the service quickly. Anchor every question to two questions — online or offline, one-time or ongoing — and the field narrows fast: Snow Family for offline bulk transfer, DataSync for online migration and scheduled sync, Storage Gateway for ongoing hybrid access (file, volume, or tape — and know cached vs. stored), and Transfer Family for managed partner file exchange over SFTP and friends, with DMS reserved for databases. Do the bandwidth math when the scenario hints at scale, memorize the cached-vs-stored cue, and these logistics puzzles become some of the most dependable points on the exam — and the same instincts you’ll use the first time you plan a real migration on the job.