AWS Solutions Architect Associate Exam Format and Time Management

Question types and exam time

AWS presents 65 questions and allows 130 minutes. Multiple-choice items have one correct answer and three distractors. Multiple-response items ask you to select two or more correct answers from at least five options. Read the direction carefully: selecting a single option on a multiple-response item can leave part of the requirement unanswered.

Fifty questions affect your score. Fifteen are unscored evaluation items and are not identified during the exam. You cannot know which items are unscored, so give every question a reasonable effort. AWS states unanswered items are scored as incorrect and there is no penalty for guessing.

The average allowance is two minutes per presented question. Some items are brief; others describe a workload, constraints, and an operational goal. Spend time understanding the requirement before comparing services. If an item is taking unusually long, choose the best answer you can, use any available review feature as explained in the exam tutorial, and continue.

Parse the architecture requirement

Architecture questions often include words that define the answer: least operational effort, lowest latency, multi-AZ recovery, predictable throughput, private access, or cost optimization. Identify the non-negotiable requirement and any secondary constraints. Then eliminate options that violate those conditions, even if they use a familiar service.

Separate the workload from the proposed implementation. A scenario may describe a relational application but not require a particular database product. Ask what consistency, query pattern, scaling behavior, availability, and recovery the workload needs. Choose the architecture that meets those requirements with the simplest suitable design.

Multiple-response questions require all selected choices to be correct. Avoid adding a plausible extra service just because it could be useful. If the stem asks for two actions, select the two that together satisfy the stated requirement. Treat every choice as a claim that must be justified by the scenario.

A practical pacing plan

1. Use the exam tutorial to understand the interface before answering.
2. Read the final sentence first if the scenario is long, then read the details to identify constraints.
3. Answer direct items promptly and reserve longer reasoning for multi-service tradeoffs.
4. Watch time at natural checkpoints and avoid spending several minutes debating two options without new evidence.
5. Use review tools only as described in the live interface; confirm that every item has an answer before submitting.

The exam page may provide service abbreviations or a help feature for shortened names. Learn to recognize common short names, but do not depend on memorizing abbreviations alone. Service names are less useful than knowing workload fit and tradeoffs.

Worked example: choose resilience without overbuilding

A business application runs across two Availability Zones and uses a managed relational database. The business needs the database to recover automatically from an Availability Zone failure. It also needs protection from accidental table deletion. Which design feature addresses the first requirement, and what separate measure addresses the second?

1. A Multi-AZ database deployment for availability, plus backups or point-in-time recovery for accidental deletion.
2. A read replica for both requirements, because every replica is a backup.
3. A larger database instance for availability, plus a public subnet for backups.
4. A load balancer alone for both requirements, because it distributes traffic and preserves data.

Best answer: A. Multi-AZ deployment supports availability through an infrastructure failure, while backup or point-in-time recovery supports restoration from logical deletion. A read replica can serve read scaling or other uses but should not be treated as the only backup. Instance size does not remove a zone failure, and a load balancer does not preserve database state.

This type of item tests whether you distinguish high availability from recovery. Both matter, but the controls are not interchangeable. In the exam, use the recovery objective and failure type to choose the service pattern.

Avoid common timing mistakes

Do not calculate the scored-question percentage during the exam. Since scored and unscored items are mixed and not labeled, every item deserves attention. Do not leave a multiple-response item blank because you are unsure; eliminate choices and make the best supported selection.

A common inefficiency is rereading the same paragraph without reducing uncertainty. Summarize the problem in one sentence: “needs a shared file system,” “must survive a zone failure,” or “requires private access with least operations.” Then evaluate each option against that statement. This is faster than mentally replaying every service detail.

Practice under the same time limit in preparation, but remember that a practice interface may differ from Pearson VUE. Use the AWS tutorial and confirmation instructions for the current interface and exam-day rules. Static practice improves reasoning; it does not reproduce the official test environment.

Use the average without forcing every question into two minutes The 65-question exam lasts 130 minutes, an average of two minutes per presented item. Some questions need less time; others include several requirements and deserve a slower read. Use a checkpoint after about 15 or 20 questions to see whether you are moving steadily. If a question is consuming too much time, identify its central constraint, eliminate options that violate it, choose the strongest remaining design, and continue. Multiple response questions require you to select more than one option. Read the instruction for how many choices are needed and verify that the selected set satisfies the whole requirement. One correct option does not make an incomplete set correct. On multiple choice questions, compare each alternative with the stated workload rather than selecting the service you recognize best. Translate architecture language into a small set of constraints: availability, recovery, latency, throughput, security, operational effort, and cost. A question may mention all of these, but the final sentence tells you which one matters most. If it asks for the lowest operational overhead, a self-managed design may be technically possible but not the best answer. If it asks for the lowest cost while preserving a firm recovery target, do not ignore the target to select the cheapest option. A pacing exercise with a constrained workload Suppose a practice section begins with an architecture question about a public website that has irregular traffic and must continue after one Availability Zone fails. First mark the availability requirement, then identify which answer uses capacity and dependencies across the needed failure scope. Next, inspect whether the database or session state remains a single point of failure. An answer that distributes web servers but leaves all data in one failed zone does not meet the whole requirement. Suppose the next question asks about long-term recovery from accidental deletion. Do not reuse the multi-zone availability solution automatically. A standby can help service continuity, but the scenario asks to restore data to an earlier point. Look for a recovery mechanism with a retained restore point and a process that can be tested. The same architecture can need separate availability and recovery controls. At a checkpoint, use remaining minutes to calculate the average time available for remaining questions. If 40 minutes remain and 25 questions are unanswered, you have 96 seconds each. That is still workable if you stop revisiting settled items. Answer every question; unanswered items are incorrect and there is no penalty for an incorrect response. During review, change an answer only when you find a concrete reason in the stem or recognize a specific mistake. A vague feeling that another option sounds more advanced is not evidence. Keep a small note of the requirement that decided the question, then move on. This preserves time for items you have not yet seen.

Use the average without forcing every question into two minutes The 65-question exam lasts 130 minutes, an average of two minutes per presented item. Some questions need less time; others include several requirements and deserve a slower read. Use a checkpoint after about 15 or 20 questions to see whether you are moving steadily. If a question is consuming too much time, identify its central constraint, eliminate options that violate it, choose the strongest remaining design, and continue. Multiple response questions require you to select more than one option. Read the instruction for how many choices are needed and verify that the selected set satisfies the whole requirement. One correct option does not make an incomplete set correct. On multiple choice questions, compare each alternative with the stated workload rather than selecting the service you recognize best. Translate architecture language into a small set of constraints: availability, recovery, latency, throughput, security, operational effort, and cost. A question may mention all of these, but the final sentence tells you which one matters most. If it asks for the lowest operational overhead, a self-managed design may be technically possible but not the best answer. If it asks for the lowest cost while preserving a firm recovery target, do not ignore the target to select the cheapest option. A pacing exercise with a constrained workload Suppose a practice section begins with an architecture question about a public website that has irregular traffic and must continue after one Availability Zone fails. First mark the availability requirement, then identify which answer uses capacity and dependencies across the needed failure scope. Next, inspect whether the database or session state remains a single point of failure. An answer that distributes web servers but leaves all data in one failed zone does not meet the whole requirement. Suppose the next question asks about long-term recovery from accidental deletion. Do not reuse the multi-zone availability solution automatically. A standby can help service continuity, but the scenario asks to restore data to an earlier point. Look for a recovery mechanism with a retained restore point and a process that can be tested. The same architecture can need separate availability and recovery controls. At a checkpoint, use remaining minutes to calculate the average time available for remaining questions. If 40 minutes remain and 25 questions are unanswered, you have 96 seconds each. That is still workable if you stop revisiting settled items. Answer every question; unanswered items are incorrect and there is no penalty for an incorrect response. During review, change an answer only when you find a concrete reason in the stem or recognize a specific mistake. A vague feeling that another option sounds more advanced is not evidence. Keep a small note of the requirement that decided the question, then move on. This preserves time for items you have not yet seen.