Q1: How do you optimize BigQuery costs?

Answer:

1. Use partitioning and clustering to reduce data scanned
2. Use materialized views for repeated queries
3. Use flat-rate pricing for consistent workloads
4. Use BI Engine for dashboards (10x faster, lower cost)
5. Avoid SELECT * - scan only needed columns
6. Use approximate aggregation functions
7. Use reserved slots for predictable workloads
8. Archive old data to Coldline storage

Q2: How do you optimize Dataflow costs?

Answer:

1. Use FlexRS for non-urgent jobs (50% savings)
2. Enable autoscaling to match workload
3. Right-size machine types (don't over-provision)
4. Minimize shuffle operations
5. Use pre-emptible workers for batch jobs
6. Optimize window sizes for streaming
7. Monitor and adjust pipeline parameters
8. Use Streaming Engine for stateful operations

Q3: How do you optimize Dataproc costs?

Answer:

1. Use pre-emptible VMs for worker nodes (91% savings)
2. Use autoscaling to match workload
3. Right-size cluster based on data volume
4. Use SSD only when needed (10x more expensive)
5. Delete clusters when not in use
6. Use preemptible VMs with auto-scaling
7. Optimize Spark configurations
8. Use regional buckets for storage

Q4: What is the GCP pricing model for compute?

Answer:

• On-demand: Pay per second, no commitment
• Preemptible/Spot: Up to 91% discount, can be terminated
• Committed use: 1-3 year commitment, 57% discount
• Sustained use: Automatic discount for long-running workloads

Use preemptible for fault-tolerant batch, committed for steady-state, on-demand for variable.

Q5: How do you implement cost monitoring?

Answer:

1. Set up Cloud Billing budgets with alerts
2. Use Cost Explorer for analysis
3. Use BigQuery cost reports
4. Set up project-level budgets
5. Use labels for cost allocation
6. Review monthly cost reports
7. Set up alerts at 50%, 75%, 100% of budget
8. Use cost anomaly detection

Q6-10: Quick-Fire Questions

Q6: What is the benefit of preemptible VMs? A: Up to 91% cost savings for fault-tolerant batch workloads. Maximum 24-hour lifetime. Can be terminated by Google. Use for batch processing and Spark jobs.

Q7: What is sustained use discount? A: Automatic discount for long-running workloads. Up to 30% for Compute Engine. Applied automatically, no commitment required. Best for steady-state workloads.

Q8: How do you estimate cloud costs? A: Use Pricing Calculator, review billing reports, use Cloud Monitoring for usage, set up budget alerts, use cost allocation labels.

Q9: What is the difference between committed and flex slots? A: Committed: 1-3 year commitment, 57% discount. Flex: 60-second commitment, can be paused. Use committed for steady-state, flex for variable.

Q10: How do you handle cost overruns? A: 1) Identify root cause, 2) Right-size resources, 3) Implement budgets, 4) Use preemptible for batch, 5) Review and optimize, 6) Set up alerts.

Q11-15: Scenario-Based Questions

Q11: Your BigQuery costs are too high. How do you optimize? A: 1) Add partitioning/clustering, 2) Use materialized views, 3) Use flat-rate for consistent workloads, 4) Avoid SELECT *, 5) Use BI Engine, 6) Archive old data.

Q12: Design a cost-effective data warehouse. A: Use BigQuery with partitioning/clustering, materialized views, BI Engine, flat-rate pricing, and Coldline for archival. Monitor with cost reports.

Q13: How do you optimize Dataflow streaming costs? A: Use Streaming Engine, optimize window sizes, use BigQuery streaming inserts efficiently, implement early triggers, use BI Engine for dashboards.

Q14: Design a cost-effective ML pipeline. A: Use Vertex AI with preemptible VMs, BigQuery ML for SQL-based ML, Dataflow for feature engineering, and Cloud Functions for inference.

Q15: How do you forecast cloud costs? A: Use historical trends, review growth rates, plan for new projects, set budgets, use Pricing Calculator, review monthly.

Q16-20: Advanced Topics

Q16: What is the difference between CapEx and OpEx in cloud? A: CapEx: Upfront hardware investment. OpEx: Pay-as-you-go. Cloud shifts CapEx to OpEx, reducing upfront costs and improving cash flow.

Q17: How do you implement showback/chargeback? A: Use labels for cost allocation, Cloud Billing reports, BigQuery for cost analysis, and export to BI tools for dashboards.

Q18: What is the benefit of committed use discounts? A: Up to 57% discount for 1-3 year commitments. Best for steady-state workloads. Reduces costs significantly for predictable workloads.

Q19: How do you optimize storage costs? A: Use lifecycle policies, implement data tiering (Standard → Nearline → Coldline → Archive), compress data, delete unnecessary data.

Q20: Design a cost governance framework. A: Implement budgets, alerts, cost allocation labels, monthly reviews, right-sizing, and optimization recommendations.

Q21-25: Cost Comparison

Q21: Compare BigQuery pricing models. A: On-demand: $5/TB scanned. Flex slots:$20/slot/month (60-second commitment). Committed slots: $40/slot/month (1-3 year commitment).

Q22: Compare storage costs across GCP services. A: GCS Standard: $0.020/GB/mo. BigQuery:$0.020/GB/mo. Persistent Disk: $0.170/GB/mo (SSD). Use GCS for data lake, BigQuery for analytics.

Q23: Compare compute costs for batch processing. A: Compute Engine on-demand: $0.0475/vCPU-hr. Preemptible:$0.016/vCPU-hr (91% savings). Dataflow: $0.08/vCPU-hr (includes management).

Q24: What is the TCO for a data warehouse migration? A: Include: migration costs, training, licensing, infrastructure, and operational costs. Cloud reduces TCO by eliminating hardware maintenance.

Q25: How do you calculate ROI for cloud migration? A: Compare: on-prem hardware, maintenance, power, cooling, staff vs. cloud costs. Include: productivity gains, faster time-to-market, reduced risk.