Understanding Serverless: Architecture, Workflow, and Observability

Definition of Serverless

Serverless is commonly defined as the combination of Backend‑as‑a‑Service (BaaS) and Functions‑as‑a‑Service (FaaS). BaaS provides managed cloud services such as databases (e.g., Parse, Firebase), authentication (e.g., Auth0, AWS Cognito) that replace custom backend components. FaaS runs developer‑written code in stateless, event‑driven containers with very short lifetimes; AWS Lambda is a typical example. Martin Fowler’s model Serverless = BaaS + FaaS emphasizes stateless containers, event‑driven execution, elastic scaling and pay‑per‑use pricing.

Serverless Development Workflow

Select a runtime offered by the FaaS provider and develop the function in a familiar language.

Test locally (most CLI tools provide emulators) and package the code.

Upload the package to the FaaS platform via CLI, console or CI pipeline.

Configure triggers – HTTP/API Gateway, cloud events, timers, etc.

The platform automatically scales the function based on concurrency and bills only for actual execution time and resources.

Example: Refactoring a Traditional Three‑Tier Web Application

A typical e‑commerce app with a Java backend, HTML/JS client, and a relational database can be transformed as follows:

Replace server‑side authentication with a BaaS provider (e.g., AWS Cognito).

Move UI‑related logic to the client, allowing direct read‑only access to data via the BaaS database.

Extract heavy or isolated tasks (search, purchase) into FaaS functions. Each function is invoked through an API gateway and can reuse the same database.

Deploy the functions, configure the gateway, and let the cloud provider handle scaling, availability and security.

After migration, developers focus solely on the business logic inside functions; infrastructure concerns such as servers, load balancers, and scaling are managed by the cloud.

Serverless Tooling

Two categories of CLI/IDE tools are widely used:

Vendor‑specific tools – e.g., AWS SAM CLI, Alibaba Cloud Function Compute funcraft. They offer deep integration with the provider’s services.

Third‑party frameworks – e.g., Serverless Framework, Serverless Devs. They support multi‑cloud deployment, YAML‑based configuration, and optional visual editors.

IDE plugins (VS Code, IntelliJ, JetBrains IDEs) enable local debugging, function packaging and one‑click deployment.

Serverless Workflow

Serverless Workflow is a fully managed orchestration service that coordinates distributed tasks using constructs such as sequences, branches and parallelism. Key features:

State tracking and checkpointing for each step.

Configurable retry and error‑handling policies.

Built‑in logging, audit and tracing.

Pay‑per‑execution‑step; no charge after workflow completion.

It allows developers to compose complex business processes across heterogeneous services (different clouds, languages, runtimes) without writing custom orchestration code.

Observability for Serverless Applications

Effective monitoring includes three dimensions:

Metrics – request count, latency, error rate, concurrent executions, memory usage.

Logs – per‑invocation stdout/stderr captured by the platform.

Tracing – distributed request IDs and call graphs for end‑to‑end latency analysis.

Alibaba Cloud Function Compute, for example, provides a console view that aggregates overall service metrics, per‑function metrics, request logs and detailed trace information. Developers can set alerts on concurrency thresholds and use the tracing view to diagnose latency spikes.