Alibaba Cloud CAP-C01 - Alibaba Cloud Certified Professional: Cloud Architect

The threats described exist at different layers, so the architecture should apply layered protection. WAF is the correct control for public HTTP/API traffic because it analyzes application-layer requests and blocks threats such as SQL injection, cross-site scripting, command injection, brute-force attacks, and other OWASP-style attacks. Alibaba Cloud explicitly supports integrating WAF with API Gateway and recommends disabling direct access paths afterward so clients cannot bypass WAF.

Large DDoS attacks require a dedicated volumetric mitigation service. Anti-DDoS Proxy scrubs malicious network and transport-layer traffic before clean traffic is forwarded toward the application infrastructure. Alibaba Cloud ' s Anti-DDoS architecture supports protected services behind Server Load Balancer resources and provides port-forwarding mechanisms for non-HTTP workloads.

Alibaba Cloud specifically recommends combining Anti-DDoS and WAF when an application needs protection against both large volumetric attacks and sophisticated application-layer exploits.

WAF should protect the HTTP/API application boundary rather than being treated as a general Layer-4 NLB firewall. Security Center provides workload security and posture management, while basic DDoS protection alone is less suitable for the question ' s explicitly stated large and sophisticated attacks.

Study Guide reference: Securing Workloads on Alibaba Cloud — WAF, Anti-DDoS, API Gateway security, and defense-in-depth.

==================================================

Tair, Auto Scaling, and ALB directly address the three core runtime requirements: session performance, elastic capacity, and distributed traffic handling.

Tair provides an in-memory Redis-compatible data tier that is appropriate for high-frequency session information. Keeping session state outside individual ECS game servers allows requests to be handled by different backend instances without losing player session information and avoids disk/database latency on critical session lookups.

Auto Scaling directly satisfies the requirement to react automatically to sudden player surges. Alibaba Cloud Auto Scaling can monitor metrics such as CPU utilization and automatically add ECS instances when capacity thresholds are exceeded, then remove excess capacity as demand decreases.

ALB distributes client requests across the resulting game-server fleet, performs health checks, and avoids overloading individual ECS instances.

CDN is highly valuable for downloading static assets such as textures, patches, images, and game packages because those objects can be cached at edge POPs. However, standard CDN does not provide the core dynamic scaling mechanism for real-time game sessions. Given that exactly three selections are required, Auto Scaling is essential because the question explicitly requires automatic scaling during player surges.

Study Guide reference: Building Highly Available, Performant Cloud Architecture — Tair, ALB, Auto Scaling, low-latency session architecture, and elastic gaming workloads.

==================================================

Serverless App Engine is the strongest fit because it provides an application-oriented serverless platform specifically intended for hosting containerized and microservice applications without requiring the organization to administer Kubernetes clusters or ECS infrastructure. SAE can deploy applications directly from container images, including images stored in Alibaba Cloud Container Registry, and automatically manages provisioning, application lifecycle, scaling, monitoring, and underlying resources.

SAE namespaces also provide logical isolation for development, testing, and production environments. This directly addresses the requirement for separate production and testing deployments while avoiding the overhead of maintaining separate ACK clusters.

SAE integrates with Application Load Balancer for Layer 7 ingress and routing. ALB can expose SAE applications and route traffic to the appropriate backend application instances.

ACK Serverless removes worker-node management but still introduces Kubernetes objects and cluster administration. ACK Managed requires even greater cluster administration. Function Compute is optimized for functions and event-driven processing rather than straightforward migration of a multi-service container application.

Study Guide reference: Core Infrastructure Deep Dive — serverless compute, SAE, containerized applications, microservices, and application scaling.

==================================================

Option B provides both an appropriate security boundary and an optimized delivery path. Alibaba Cloud CDN should be the Internet-facing content-delivery layer so that cacheable HTTPS resources are served from edge points of presence close to users. Alibaba Cloud currently supports directly configuring an ALB instance domain as a CDN origin, producing the request path client → CDN → ALB → backend server. This prevents clients from bypassing the load-balancing tier and preserves centralized health checking and traffic distribution.

The backend ECS instances should remain on private network addresses rather than being individually Internet-exposed. For an Internet-facing ALB, external traffic reaches the ALB through its public EIP and is then forwarded internally to backend ECS private IP addresses. This substantially reduces the attack surface while retaining public application availability.

Options C and D unnecessarily expose ECS instances through public networking. Option A uses private ECS instances, but making the individual application servers the CDN origins bypasses the ALB ' s centralized distribution and availability functions. Therefore, CDN → ALB → private ECS is the strongest architecture.

Study Guide reference: Delivering Content on Alibaba Cloud; Delivering Services on Alibaba Cloud; Highly Available Architecture.

==================================================

Function Compute with EventBridge provides the lowest-management architecture among the choices. Function Compute is a fully managed, event-driven compute platform: Alibaba Cloud provisions and scales the required execution infrastructure automatically, and billing is based on configured resources and execution duration. This removes the operational burden of maintaining ECS instances or an ACK cluster.

The stated execution time of up to one hour is supported. Current Function Compute documentation permits individual function executions of up to 24 hours, so the job duration does not require a container cluster merely because it exceeds traditional short-function runtimes.

EventBridge supports recurring time-triggered events using fixed schedules or cron expressions and is suitable for scheduled batch-processing workflows. Function Compute also integrates directly with EventBridge-based cloud-product event triggers.

Options A and C introduce Kubernetes control and workload-management complexity that is unnecessary for a single scheduled processing job. Option B is less precise because Serverless Workflow provides orchestration rather than being required to perform the actual computation.

Therefore, EventBridge schedules the daily invocation and Function Compute performs the OSS-backed processing without persistent infrastructure.

Study Guide reference: Delivering Services on Alibaba Cloud; serverless architecture and event-driven application delivery.

==================================================

Data Management Service is specifically designed to centralize database administration, authorization, and governance. Its permission model provides fine-grained control over instances, databases, tables, columns, rows, and sensitive data. Administrators can control logon, query, export, and change privileges while maintaining centralized governance rather than manually coordinating several independent services.

DMS is also appropriate where the organization needs auditable database access. Alibaba Cloud explicitly describes DMS authorization scenarios that include controlling RDS access and auditing employee data access.

RAM controls permissions to Alibaba Cloud resources and APIs, but combining RAM with ActionTrail would require multiple mechanisms and would not provide DMS ' s database-object-level permission model. An RDS IP whitelist is useful as a network access restriction but cannot express granular user permissions or provide centralized database governance. WAF protects HTTP/HTTPS applications and is not the appropriate control plane for direct RDS database authorization.

DMS therefore provides the required access control and audit functionality with the lowest operational complexity by consolidating database security administration into a purpose-built managed service.

Study Guide reference: Securing Workloads on Alibaba Cloud — DMS access control, database governance, RDS security, and auditing.

Global Database Network is the architecture specifically intended for geographically distributed PolarDB deployments requiring cross-region data replication, local read capability, and disaster recovery. A GDN contains a primary cluster plus secondary clusters in other regions. Data is continuously synchronized between them. Applications can connect to regional cluster endpoints; read requests can be served locally while write requests are routed to the primary cluster. Alibaba Cloud also supports switching the primary cluster to a secondary cluster, including forced switchover scenarios for failures.

Option C provides strong node-level and zone-level availability within a PolarDB cluster, but it does not deliver the same level of protection against an entire regional outage. Option D relies on backup restoration, which introduces significantly higher RTO and operational intervention. Option B invents an unnecessary Function Compute replication mechanism instead of using PolarDB ' s native cross-region replication.

For a mission-critical global e-commerce database, GDN provides the best combination of cross-region redundancy, replication, local reads, and regional failover capabilities.

Study Guide reference: Building Highly Available, Performant Cloud Architecture — PolarDB high availability, cross-region disaster recovery, and Global Database Network.

The existing architecture has two important bottlenecks: a single ECS compute node and files stored on storage attached to that individual ECS instance. Option E eliminates the compute single point of failure. A custom image allows identical ECS instances to be created, while Auto Scaling maintains multiple instances and ALB distributes requests across healthy backends. This provides horizontal scalability and significantly stronger application availability.

Option D addresses shared application storage. File Storage NAS is a managed, POSIX-compatible shared filesystem that multiple ECS instances can mount concurrently. All web servers therefore see the same files, while NAS handles storage scalability and consistency independently from the ECS lifecycle.

Option A improves performance by moving cacheable website content toward users. CDN serves static resources from nearby POPs and reduces requests and bandwidth reaching the origin infrastructure.

Although OSS is excellent for native object-storage workloads, mounting OSS through ossfs does not provide the same filesystem semantics and consistency characteristics as NAS for a traditional application requiring minimal modification. Global Accelerator primarily optimizes network paths and does not eliminate the application ' s compute or storage single points of failure.

Study Guide reference: Building Highly Available, Performant Cloud Architecture — ALB, Auto Scaling, NAS shared storage, and CDN.

==================================================

Tair (Redis OSS-Compatible) is the appropriate centralized session-state layer for an elastic, distributed application. When ECS instances frequently scale in and out, application sessions must not depend on the lifecycle of any individual compute node. By storing session tokens and session state in Tair, every ECS instance can retrieve the same session information, making the web/application tier effectively stateless and compatible with Auto Scaling.

Alibaba Cloud Tair is a fully managed Redis-compatible database designed for high throughput, low latency, elastic scalability, and high availability. Alibaba Cloud also documents session-oriented Tair use cases in which session tokens are stored as Redis/Tair data structures with TTL values, demonstrating precisely why the service fits distributed session management.

Session stickiness is weaker because it attempts to keep a user on a specific backend and therefore creates affinity to an instance that might be removed during scaling or failover. CloudMonitor is an observability service, not a session store. STS issues temporary Alibaba Cloud credentials and does not maintain application user sessions.

Consequently, separating session state from ECS compute by storing it in Tair gives the architecture the elasticity and fault tolerance required.

Study Guide reference: Distributed Architecture on Alibaba Cloud; Performant Architecture; Core Infrastructure Deep Dive.

==================================================