Examkin

Soru

Zorluk: ZorDesigning Infrastructure for Technical Requirements and High Availability

An online multiplayer gaming platform is architecting its global matchmaking and player profile backend on Google Cloud. The architecture must achieve high availability across regional outages with a Recovery Point Objective (RPO) of 0 and a Recovery Time Objective (RTO) of less than 1 minute for transactional player state. Additionally, the compute tier handles persistent WebSocket connections and I/O-intensive queue processing, which does not correlate directly with CPU usage. Which TWO architectural decisions should the Cloud Architect implement to satisfy these technical requirements? (Select 2 answers)

  1. Deploy a multi-region Cloud Spanner instance to provide globally distributed strongly consistent transactions with zero RPO across regional failures.Cevap
  2. B
    Deploy a regional Cloud SQL for PostgreSQL instance with asynchronous cross-region read replicas and manual failover procedures.
  3. Configure Kubernetes Horizontal Pod Autoscaler (HPA) using custom metrics for active WebSocket connections and queue length rather than default CPU utilization.Cevap
  4. D
    Configure Kubernetes Horizontal Pod Autoscaler (HPA) based exclusively on average CPU utilization target of 80% across pods.
  5. E
    Connect the compute VPC to a central hub VPC using VPC Network Peering and rely on transitive routing to access external third-party services.

Cevap

The architect should deploy a multi-region Cloud Spanner instance for zero-RPO transactional data and configure Kubernetes Horizontal Pod Autoscaler using custom metrics for active connections and queue length.
Deploying a multi-region Cloud Spanner instance guarantees synchronous multi-region replication with zero data loss (RPO = 0) and automatic sub-minute failover. Additionally, configuring HPA with custom application metrics (WebSocket connections and queue depth) ensures the compute tier scales appropriately for I/O-bound workloads that do not correlate with CPU consumption.

Adım Adım Çözüm

1
Evaluate storage tier against RPO=0 and sub-minute RTO requirements during regional failures.
Multi-region Cloud Spanner provides synchronous replication across regions, meeting RPO=0 and sub-minute failover, whereas asynchronous Cloud SQL replicas risk data loss and slower failover.
Cloud Spanner uses Paxos consensus across multi-region zones to guarantee strong consistency with minimal recovery objectives.
2
Evaluate autoscaling criteria for persistent connection and queue-bound microservices.
Scaling based on custom metrics (active WebSocket connections and queue depth) accurately matches workload demand, whereas CPU metrics lag or misjudge I/O bottlenecks.
Connection-heavy workloads consume memory and network sockets without spiking CPU, making CPU-based autoscaling ineffective.

Anahtar Kavram

Designing multi-region database architectures for zero RPO/RTO and autoscaling I/O-bound compute workloads using custom metrics.
Bu soruyu puanla

Konular

Designing Infrastructure for Technical Requirements and High Availability