AWS Certified Solutions Architect – Associate remains the current SAA-C03 exam, and the official guide still defines the role around designing solutions that are secure, resilient, high-performing, and cost-optimized. AWS expects candidates to translate business requirements into architectures, review existing systems, and identify improvements rather than simply recognize service names.
The current Amazon SAA-C03 blueprint has four scored domains: Design Secure Architectures at 30%, Design Resilient Architectures at 26%, Design High-Performing Architectures at 24%, and Design Cost-Optimized Architectures at 20%. AWS lists a 130-minute exam with 65 questions, of which 50 are scored and 15 are unscored, and a $150 USD registration fee. The minimum scaled passing score is 720.
The exam is built around the AWS Well-Architected way of thinking
The target candidate is a solutions architect with at least one year of hands-on experience designing AWS cloud solutions. That experience matters because scenario questions often provide several technically possible designs and ask for the option that best satisfies security, reliability, performance, operational simplicity, or cost requirements.
Within the broader Amazon certification catalog, this credential is a general architecture exam rather than a deep operations or specialty exam. Candidates should know the major AWS service families well enough to choose among them, but service selection is always tied to a requirement.
Secure architectures begin with identity, network boundaries, and data protection
Domain 1 covers secure access to AWS resources, secure workloads and applications, and appropriate data-security controls. That means IAM policies and roles, Organizations, identity federation, encryption, key management, Secrets Manager, network segmentation, security groups, network ACLs, VPC endpoints, and service-level security can all appear in architecture decisions.
A review of security groups and network ACLs can reinforce stateful-versus-stateless filtering, while Amazon S3 server-side encryption provides a concrete data-protection example. The exam expects candidates to apply least privilege and protect data in transit and at rest without adding unnecessary administrative complexity.
Resilient architectures separate failure domains and remove tight coupling
Domain 2 includes scalable and loosely coupled architectures plus highly available and fault-tolerant design. Candidates should understand multiple Availability Zones, load balancing, Auto Scaling, queues, publish/subscribe, serverless patterns, microservices, caching, edge delivery, and disaster-recovery choices.
The key idea is that one component’s failure should not unnecessarily take down the entire system. Amazon SQS is a classic decoupling tool, while EC2 Auto Scaling can add or replace compute capacity as demand or health changes. The right architecture also considers state, retries, idempotence, and what happens when dependencies are temporarily unavailable.
High performance is not the same as choosing the largest instance
Domain 3 covers high-performing storage, compute, database, network, and data-ingestion solutions. Candidates should compare S3 storage patterns, EBS and EFS, EC2 families, Lambda and containers, RDS and Aurora, DynamoDB, caching, CloudFront, Route 53, Direct Connect, VPN, Kinesis, Glue, and other in-scope services based on workload behavior.
For example, RDS Multi-AZ and read replicas solve different problems: one primarily improves availability while the other can scale reads and support other use cases. Similarly, Lambda and EC2 have different operational and scaling models. The exam rewards matching architecture to demand rather than memorizing a universal performance hierarchy.
Cost optimization changes architecture choices across every layer
Domain 4 covers storage, compute, database, and network cost optimization. Candidates should understand access patterns, storage tiers, lifecycle policies, EC2 purchasing options, serverless economics, managed-service trade-offs, database engines, data transfer, NAT architecture, VPC endpoints, content delivery, and AWS cost-management tools.
S3 lifecycle policies and S3 Intelligent-Tiering are useful examples because storage cost depends on access frequency and retention. Network design matters too: unnecessary cross-AZ or cross-Region traffic can create recurring cost even when the application works perfectly.
VPC architecture is the common network foundation behind many scenarios
Subnets, route tables, internet gateways, NAT gateways, security groups, network ACLs, VPC endpoints, peering, Transit Gateway, VPN, and Direct Connect influence both security and performance. Candidates should know which resources need public reachability, which should remain private, and how traffic exits or crosses network boundaries.
A review of Amazon VPC fundamentals can help connect these components. The exam usually cares more about the resulting architecture than about low-level CLI syntax: public versus private access, highly available egress, private service connectivity, and scalable multi-VPC design.
Storage questions are really questions about access pattern, durability, and lifecycle
Amazon S3, EBS, EFS, FSx families, Storage Gateway, Backup, and archival services address different access and durability requirements. A candidate should identify whether the workload needs object, block, or shared file storage, how frequently the data is accessed, how fast it must be restored, and how long it must be retained.
Amazon S3 is especially important because it appears across backup, static content, data lakes, lifecycle, replication, encryption, and event-driven workflows. The strongest answer usually follows the workload’s access semantics rather than the popularity of the service.
Database design is a choice among access models and operational responsibilities
Relational workloads may fit RDS or Aurora; key-value workloads may fit DynamoDB; analytics, graph, document, caching, time-series, and other use cases have additional managed choices. Candidates should reason about consistency, transactions, scale, latency, availability, and management overhead.
A review of Amazon RDS and DynamoDB-style NoSQL trade-offs can strengthen this area. The exam does not ask candidates to choose a database based on brand familiarity. It asks which data model and operating model best satisfy the stated application.
Edge, DNS, and event-driven services connect performance with resilience
CloudFront can cache content closer to users, Route 53 can route DNS queries based on health or routing policy, API Gateway can front APIs, and event-driven services can reduce synchronous coupling. These services often improve more than one Well-Architected concern at the same time.
CloudFront and Route 53 are worth studying together because global applications frequently combine DNS routing, health checks, and edge caching. The correct combination depends on whether the requirement is latency, failover, geographic routing, or content delivery.
The live exam rewards trade-off reasoning more than service memorization
The current AWS guide publishes a broad, non-exhaustive list of in-scope services and warns that the list can change. That makes service-by-service memorization fragile. A better readiness test is to take a workload and explain its identity, network, storage, compute, database, availability, performance, backup, monitoring, and cost decisions.
AWS also publishes the exam as compensatory: candidates do not need to pass each domain separately, but the overall scaled score must meet the passing standard. That does not mean a weak domain can be ignored. Security carries the largest weighting, and the other three domains collectively contain many scenarios where one architecture decision influences several concerns. Balanced preparation remains the safest approach because scored questions are distributed across all four areas.
The in-scope service list is intentionally non-exhaustive and subject to change. That is an important preparation signal. Candidates should learn service families and architectural patterns deeply enough to reason about a service even when its exact feature was not emphasized in older notes. AWS can also use official short service names on the exam, so familiarity with common abbreviations and the function of each service matters more than memorizing long formal names.
Security scenarios often combine resource policies, identity policies, encryption, and network controls. For example, an S3 bucket can be privately reachable and encrypted while still being inaccessible because the IAM principal lacks permission, or an authorized principal can be blocked by a restrictive resource policy. The strongest candidates separate authentication, authorization, network reachability, and data protection instead of treating them as one security setting.
Resilience also includes designing for demand spikes and dependency failure, not only hardware failure. A multi-AZ deployment can survive an infrastructure event but still collapse if every request synchronously waits on one overloaded downstream service. Queues, caches, replicas, and asynchronous workflows are resilience tools because they change how components react to pressure and partial failure.
Performance questions should be translated into a measurable bottleneck: storage throughput, IOPS, network latency, database read capacity, compute concurrency, cache hit rate, or data-ingestion rate. Once the limiting resource is clear, the service choice becomes easier. “High performance” is not a service tier; it is an architecture that matches the workload’s dominant constraint.
Cost questions frequently hide in topology. NAT gateways, cross-AZ transfer, inter-Region replication, idle load balancers, overprovisioned databases, excessive snapshots, and inefficient storage classes can dominate spend even when compute is right-sized. Architects should follow data movement and resource lifetime as carefully as instance price.
The exam’s 65-question format also matters strategically because only 50 questions are scored and the 15 unscored items are not identified. Candidates cannot safely guess which questions “do not count.” Treat every item as scored, manage time across the full 130 minutes, and use architectural elimination when two choices look plausible.
AWS states that there is no penalty for guessing, so unanswered questions should be avoided. More importantly, the distractors are designed to be plausible choices for someone with incomplete knowledge. That means preparation should focus on why the nearest alternative is weaker: wrong failure domain, unnecessary operations, excessive cost, insufficient security, or a service that solves a different performance problem.
Use the in-scope service list as a breadth check, not a study order. The architect role is expected to recognize services across analytics, application integration, compute, containers, databases, management, networking, security, storage, and migration. But the exam guide itself organizes the work by design outcomes. Outcome-first preparation remains more durable as AWS adds or renames services.
That distinction is central to current exam readiness.
That requirement-first discipline is the safest way to prepare.
The SAA-C03 preparation material can help consolidate the breadth, but final review should focus on why one architecture is preferable under a given constraint. That is the central skill the current exam is designed to validate.