SAA-C03 Exam Dumps : AWS Certified Solutions Architect - Associate (SAA-C03)

Option Censures dynamic scaling based on demand using a target tracking scaling policy, optimizing costs.

Option Aresults in over-provisioning, leading to higher costs.

Option Bincreases costs by using larger instances.

Option Dis not feasible as instance store volumes are ephemeral and unsuitable for shared storage like EFS.

Amazon SQS is the AWS-recommended service for handling decoupled, scalable, ordered, durable message ingestion with potentially large workloads. It supports payloads up to 256 KB directly and up to several MB using S3-backed large-payload extensions.

Lambda functions can process messages from SQS with automatic scaling. ECS with the Fargate launch type provides a serverless container platform for the frontend, scaling based on demand without managing servers.

SNS (Options B and D) is not a queue, does not provide durable message buffering, and is not appropriate for variable, bursty workloads that may exceed backend throughput. Lambda@Edge (Options A and D) is unsuitable for backend order processing because it runs at CloudFront edge locations with strict limitations.

Sporadic HTTP 500 errors in an application that depends on DynamoDB often indicate intermittent downstream failures (for example, transient throttling, temporary network issues, request timeouts, or service-side retryable exceptions). The prompt notes that the DynamoDB table is in on-demand mode and other applications use it without problems, which suggests the issue is not a sustained capacity shortfall or a systemic DynamoDB outage. Instead, the web application likely needs to handle occasional retryable failures more gracefully.

Option C is the best practice for improving reliability when calling AWS services: implement exponential backoff with retries. AWS SDKs commonly include retry behavior, but many systems still require tuning (for example, increasing maximum retries, adding jitter, ensuring idempotent operations where required, and retrying only on retryable errors). Exponential backoff reduces the chance of overwhelming a service during brief contention periods and helps the application recover quickly from transient errors without user-visible failures. This approach is also minimally disruptive because it is a configuration and logic improvement rather than an architectural migration.

Option A is not a standard DynamoDB configuration; DynamoDB has item size limits and throughput is managed via on-demand or provisioned capacity, not by “supporting larger write requests.” Option B (Streams) provides a change data capture feed but does not prevent or mitigate request failures; it is for event processing, replication, or auditing. Option D (strongly consistent reads) addresses stale reads, not HTTP 500 errors, and could increase read cost/RCU consumption without solving the root cause.

Therefore, adding exponential backoff and retries is the most appropriate and operationally sound way to reduce sporadic failures without disrupting the application’s architecture or performance.