3.1.2.3. Serverless Architectures for Scalability

3.1.2.3. Serverless Architectures for Scalability

Serverless eliminates capacity planning — the platform scales to zero when idle and handles thousands of concurrent requests without configuration.

AWS serverless compute:

• Lambda: Functions scale automatically (up to 1,000 concurrent by default, can request increase to tens of thousands). Pay per invocation and execution duration.
• Fargate: Container workloads without managing EC2 instances. ECS/EKS scheduler places tasks on managed infrastructure.

Serverless supporting services:

• API Gateway: Routes HTTP requests to Lambda, ECS, or other backends. Handles authentication, rate limiting, caching.
• DynamoDB: On-demand mode scales read/write capacity automatically. No table management.
• S3: Unlimited object storage with built-in scaling.
• SQS/SNS/EventBridge: Fully managed messaging with no capacity planning.

Serverless scaling characteristics:

Provisioned concurrency pre-initializes Lambda execution environments to eliminate cold starts. Use for latency-sensitive APIs where cold start delays are unacceptable.

Exam Trap: Lambda concurrent execution limits are per-region per-account. If one function consumes all 1,000 concurrent executions, other functions in the same account/region are throttled. Use reserved concurrency to guarantee capacity for critical functions and prevent noisy-neighbor problems.

Lambda + relational databases: every concurrent execution environment opens its own database connection, and environments don't share connection pools — so 1,000 concurrent invocations can exhaust max_connections. RDS Proxy pools and multiplexes those connections, avoiding both a bigger database and a concurrency cap that would throttle the function.