Amazon EFS performance and storage classes: SAA-C03

EFS is already chosen — now the exam asks you to tune it. Throughput modes, performance modes, storage-class lifecycle and mount targets, in the SAA-C03's own scenario style. Seven scenarios, 2026 edition.

Try 7 free EFS tuning scenarios in the exam's own style — every answer explained, no sign-up.

Straight to the 7 free questions ↓

Once a scenario has settled on Amazon EFS, the SAA-C03 stops asking which storage service and starts asking how to tune it. The distractors are no longer S3 or EBS — they are other EFS settings, each plausible, only one matching the requirement. The two axes the exam turns on are throughput mode and performance mode, with storage-class lifecycle and mount targets rounding out the common scenarios. The diagram lays those axes out side by side.

Throughput modes: match the pattern

EFS offers three throughput modes, and the requirement sentence picks one. Bursting ties baseline throughput to how much data is stored and earns burst credits for short spikes — fine for a general file system that grows over time, but a small file system gets a small baseline. Provisioned sets a fixed throughput independent of size, so it is the answer when a few gigabytes must sustain heavy, steady throughput. Elastic scales throughput up and down automatically and bills on what is used — the answer for spiky or unknown patterns where you do not want to size anything at all. The recurring trap is offering Provisioned at the peak (wasteful) where Elastic fits, or Bursting where the dataset is too small to burst enough.

Performance modes: latency versus parallelism

The performance mode is a different lever. General Purpose is the low-latency default and the right call for almost everything. Max I/O trades slightly higher per-operation latency for far greater aggregate throughput and operations across very large numbers of concurrent clients — the pick when thousands of instances mount one file system and total throughput outweighs single-op speed. When a stem stresses massive parallelism, Max I/O is in play; otherwise General Purpose wins. Note that a throughput mode and a performance mode are independent choices — the exam plants one where the other belongs to see if you conflate them.

Storage classes and lifecycle

Cost tuning lives in the storage classes. Standard is the Regional default, stored redundantly across multiple Availability Zones. Infrequent Access and Archive are progressively cheaper tiers for colder files, and lifecycle management transitions files to them automatically after a set idle window — all without changing how applications mount the file system. For re-creatable data where AZ-level redundancy is not needed, One Zone keeps data in a single Availability Zone at a lower price. The lever you choose depends on how the data is accessed and how much redundancy it truly needs.

Scaling and mount targets

Two facts settle several scenarios outright. EFS scales storage automatically and is multi-AZ by design on the Standard class, so there is no capacity to pre-provision and no manual standby to build for AZ resilience — an answer that proposes reserving peak capacity is describing EBS, not EFS. And instances reach the file system through a mount target in their own Availability Zone: if one AZ has no mount target, instances there simply cannot connect. Create one mount target per AZ in use, and reachability problems from a single zone disappear.

How to use the seven scenarios

Answer each before opening the explanation, then find the requirement word that decided it — “unpredictable”, “small dataset”, “thousands of clients”, “cheapest”, “cold files”, “one AZ can’t connect”. Every explanation names that constraint and takes the distractors apart, because the miss is almost always a real EFS setting answering a different question than the one asked. For the wider domain, see Design High-Performing Architectures; to revisit when EFS is the right service in the first place, see EBS vs S3 vs EFS.

A two-axis map of Amazon EFS tuning. The horizontal axis is throughput mode: Bursting, where throughput scales with stored size and earns burst credits; Provisioned, where a fixed throughput is set independently of size for a small but busy file system; and Elastic, which scales throughput up and down automatically for spiky or unknown patterns. The vertical axis is performance mode: General Purpose, the low-latency default, and Max I/O, which trades higher per-operation latency for greater parallel throughput across thousands of clients. A side note shows lifecycle management transitioning cold files from the Standard class to Infrequent Access and then to Archive to cut cost, while the file system stays mounted.

7 free SAA-C03 practice questions

Answers and explanations — no email wall

SAA-C03 Question 1 of 7

A shared Amazon EFS file system serves a batch pipeline whose throughput swings from near-idle to very heavy at unpredictable times. The team does not want to size or manage a throughput setting, and wants to pay only for what is used. Which throughput mode fits?

Answer: A — Elastic throughput, which scales performance up and down automatically to match demand.

Elastic throughput scales instantly with demand and bills on throughput actually used, so an unpredictable, spiky workload needs no manual sizing and pays nothing for idle capacity. Why not the others: Provisioned at the peak pays for that ceiling around the clock, most of it wasted between spikes. Bursting ties sustained throughput to stored size and can exhaust its burst credits under prolonged heavy use. Max I/O is a performance mode, not a throughput mode — it changes latency and parallelism, not how throughput is provisioned. Unknown or bursty demand with no sizing effort is Elastic throughput.

SAA-C03 Question 2 of 7

A genomics platform mounts one EFS file system from thousands of EC2 instances running a highly parallel analysis, and aggregate throughput matters far more than the latency of any single file operation. Which performance mode should a solutions architect choose?

Answer: A — Max I/O, which sustains higher aggregate throughput across very large numbers of concurrent clients.

Max I/O scales to higher levels of aggregate throughput and operations by relaxing the latency of each operation slightly — the right trade when massive parallelism matters more than per-op speed. Why not the others: General Purpose is the low-latency default but can cap out on operations at extreme client counts, which is exactly this case. Provisioned throughput is a throughput mode and does not change the parallelism ceiling. One Zone is a storage class about durability scope and cost, not concurrency. Thousands of clients, throughput over single-op latency, is Max I/O.

SAA-C03 Question 3 of 7

An EFS file system holds only a few gigabytes but must sustain high, steady throughput for a busy metadata service. On Bursting mode it keeps running out of throughput. What should a solutions architect recommend?

Answer: A — Switch to Provisioned throughput and set the throughput the workload needs, independent of the stored size.

In Bursting mode, baseline throughput is proportional to how much data is stored, so a tiny file system gets a tiny baseline no matter how busy it is. Provisioned throughput decouples throughput from size — you set the level you need. Why not the others: padding the file system with junk data to buy throughput is a cost and hygiene anti-pattern the exam rejects. Infrequent Access lowers storage cost for cold data and does nothing for throughput. Max I/O changes the parallelism trade-off, not the size-linked baseline. High steady throughput on a small dataset is Provisioned throughput.

SAA-C03 Question 4 of 7

A shared file system accumulates millions of files that are read heavily for a few weeks and then almost never touched, yet must remain in the same file system. The team wants to cut storage cost automatically without changing how applications mount it. What should they configure?

Answer: A — EFS lifecycle policies that transition cold files to Infrequent Access and then to the Archive class.

EFS lifecycle management moves files that have not been accessed within a set window to Infrequent Access, and colder still to Archive, cutting storage cost transparently — the file system stays mounted and the paths are unchanged. Why not the others: hand-rolling a copy-to-Glacier job leaves EFS, breaking the requirement to keep files in the same file system, and adds operational overhead. Provisioned throughput and Max I/O are performance controls with no bearing on storage-class pricing. Automatic, in-place cost reduction for cold files is EFS lifecycle to IA and Archive.

SAA-C03 Question 5 of 7

A team about to adopt EFS is drafting a capacity plan: how much storage to pre-provision and how to keep the file system available if an Availability Zone fails. What should a solutions architect tell them about the EFS Standard class?

Answer: A — EFS grows and shrinks automatically and stores data redundantly across multiple Availability Zones in the Region — no capacity to pre-provision.

EFS Standard is a Regional, elastic file system: it scales storage automatically as files are added and removed, and it stores data redundantly across multiple Availability Zones — so there is no capacity to plan and no manual standby for AZ resilience. Why not the others: pre-provisioning peak capacity describes EBS or a self-managed server, not EFS. The single-AZ description is the One Zone class, not Standard. Mount targets are network entry points, not capacity reservations. Automatic scaling and built-in multi-AZ durability is the EFS Standard baseline.

SAA-C03 Question 6 of 7

EC2 instances in Availability Zone C cannot reach an EFS file system, while instances in zones A and B mount it fine. The file system uses the Standard class. What is the most likely fix?

Answer: A — Create a mount target for the file system in Availability Zone C.

Instances connect to EFS through a mount target in their own Availability Zone; if an AZ has no mount target, instances there have no path to the file system. Creating one mount target per AZ in use resolves it. Why not the others: a Standard file system is already Regional, so it need not be recreated — only the per-AZ mount target is missing. Throughput and performance modes govern speed and parallelism, not network reachability from a zone. Missing connectivity from one AZ is a missing mount target.

SAA-C03 Question 7 of 7

A workload stores large intermediate results that can be regenerated at any time if lost, and the team wants the cheapest EFS option, accepting reduced redundancy. Which choice fits, and what is the trade-off?

Answer: A — EFS One Zone storage class, which lowers cost by keeping data in a single Availability Zone instead of across several.

EFS One Zone stores data in a single Availability Zone at a lower price than the Regional Standard class — the right pick for re-creatable data where losing the AZ-level redundancy is acceptable. Why not the others: Standard with minimal Provisioned throughput is still Regional and priced accordingly; lowering throughput does not change the storage-class cost. Lifecycle to Archive reduces cost for cold files but keeps the redundancy scope of its class rather than dropping to one AZ. Max I/O is a performance mode, not a pricing tier. Cheapest for re-creatable data, trading multi-AZ redundancy, is One Zone.

That is exactly how every question in the course works — answer, explanation, why-not. The real set continues in the practice player: 10 free questions, no sign-up.

Continue with the 10 free questions →

Those 7 questions were the start.

The exam does not test whether you recognise a term — it tests whether you can rule out three plausible answers under time pressure. That is what the explanations above are for, and there are 780 more questions built exactly like them.

Collecting questions yourself

Scattered across forums, of unknown age
Answer keys without reasoning
No idea which domain you are weak in

Practising with a system

780 questions in 12 full tests, SAA-C03 (2026)
Every option explained — including the wrong ones
Readiness per exam domain, and drills for your weakest

Start free — 10 questions

Straight into the player. No account, no email.

Frequently asked questions

Bursting or Provisioned throughput — which does the exam expect?

Bursting ties throughput to how much data is stored and suits general file systems that grow over time. Provisioned sets a fixed throughput independent of size, so it is the answer when a small file system must sustain high throughput. If the pattern is spiky or unknown and you do not want to size anything, Elastic scales automatically and bills on usage. Match the requirement word — steady on a small dataset, or unpredictable and hands-off — to the mode.

What is the difference between General Purpose and Max I/O performance mode?

General Purpose is the low-latency default and is right for the large majority of workloads. Max I/O trades slightly higher per-operation latency for much greater aggregate throughput and operations across very large numbers of concurrent clients — think thousands of instances on one file system. Choose Max I/O only when massive parallelism outweighs single-operation latency.

Are these real exam questions?

No. They are original scenarios written for this page in the exam's style and difficulty. Reproducing real exam items violates the AWS certification agreement and can cost a candidate their certification — ours come with the reasoning attached instead.

Updated for SAA-C03 (July 2026). The sample questions above are our own work in the style of the exam — not real exam items. The exam itself is set and marked by the certification body.