Examkin

Soru

Zorluk: ZorSelecting and Designing Compute and Application Platforms

A global fintech firm is designing the compute architecture on Google Cloud for a payment processing system consisting of two distinct workloads. Workload 1 is a stateless microservice that receives gRPC requests from external gateways, experiences unpredictable bursty traffic with periods of complete idle time, and requires automatic scale-to-zero capability with minimal operational overhead. Workload 2 is a stateful stream-enrichment engine that requires ultra-low latency access to local NVMe storage and custom Linux kernel tuning (`sysctl` network parameter modifications) at the operating system level. Which TWO compute platform strategies should you select to meet these technical requirements while adhering to Google Cloud best practices?

  1. Deploy Workload 1 on Cloud Run configured with end-to-end HTTP/2 (gRPC) support.Cevap
  2. Deploy Workload 2 on Compute Engine instances managed by a Managed Instance Group (MIG) with attached Local SSDs.Cevap
  3. C
    Deploy Workload 1 on a dedicated Google Kubernetes Engine (GKE) Autopilot cluster.
  4. D
    Deploy Workload 2 on Cloud Run using container images built with custom kernel flags and mounted Cloud Storage FUSE volumes.
  5. E
    Deploy both Workloads on App Engine Flexible environment custom runtimes.

Cevap

Select Cloud Run with gRPC support for Workload 1, and Compute Engine Managed Instance Groups with Local SSDs for Workload 2.
Cloud Run provides serverless scale-to-zero capabilities with native gRPC/HTTP2 support for stateless microservices, satisfying Workload 1. Compute Engine virtual machines managed via MIGs provide full administrative control to configure host OS kernel sysctl parameters and attach Local SSD NVMe disks, satisfying Workload 2.

Adım Adım Çözüm

1
Analyze Workload 1 requirements
Workload 1 is stateless, containerized, uses gRPC, has unpredictable bursty traffic, and requires scale-to-zero with minimum management overhead.
Cloud Run natively supports containerized gRPC applications, scales down to zero instances when idle to minimize costs, and removes infrastructure management responsibility.
2
Analyze Workload 2 requirements
Workload 2 requires custom kernel tuning (`sysctl` network modifications) and ultra-low latency local NVMe storage.
Serverless containers (Cloud Run / App Engine) do not permit host kernel modifications. Compute Engine virtual machines provide root-level access for OS customization and direct attachment of Local SSD (NVMe) drives.
3
Combine compute platforms according to workload characteristics
Pairing Cloud Run for the stateless microservice with Compute Engine MIGs for the OS-customized stateful workload delivers optimal performance and cost efficiency.
Matching each service to its specific operational constraint avoids both compute over-provisioning and platform capability limitations.

Anahtar Kavram

Compute Platform Selection Strategy based on Statefulness, Control, and Scaling Constraints
Bu soruyu puanla

Konular

Selecting and Designing Compute and Application Platforms