A financial analytics firm executes batch Monte Carlo risk simulations twice per week. Each execution requires rapidly scaling up to 4,000 vCPUs for approximately two hours to perform stateless calculations, after which resource utilization drops back to zero. The firm requires a solution that minimizes total compute costs, eliminates idle infrastructure overhead, and avoids maintaining control plane management overhead when simulations are not running. Which architecture should the Cloud Architect recommend?
- Containerize the simulation application and execute it using Cloud Run jobs, allowing tasks to scale on demand and automatically terminate upon completion.Answer
- BDeploy a multi-zone Google Kubernetes Engine (GKE) cluster with Cluster Autoscaler to scale worker nodes up from zero for each batch execution.
- CPurchase 3-year Compute Engine vCPU Committed Use Discounts (CUDs) sized for 4,000 vCPUs and run the jobs on a Managed Instance Group.
- DStore simulation parameters in a multi-region Cloud Spanner instance and orchestrate task execution using Cloud Functions.
Answer
Containerize the simulation application and execute it using Cloud Run jobs, allowing tasks to scale on demand and automatically terminate upon completion.
Executing the batch simulations using Cloud Run jobs is the optimal solution because Cloud Run jobs natively supports containerized, run-to-completion workloads with zero ongoing infrastructure costs when idle. It automatically provisions resources to scale out tasks and terminates immediately upon job completion, aligning cost directly with actual usage.
Step-by-Step Solution
Key Concept
Serverless Batch Execution for Intermittent Cost-Optimized Workloads
Estimated Time:2m 0s