Skip to main content

Skills application solution

Instruction iconInstruction and application
Complete iconIn Progress

Skills application solution​

Compare your skills application output to the solution example below that Multiverse experts have provided. Solution1. Initial resource requirements (inference): -Concurrent inference requests: Peak traffic = 1,000 queries/second. Inference time = 100ms = 0.1 seconds. Requests processed per core per second = 1 ÷ 0.1 = 10 requests/sec. Required cores = 1,000 ÷ 10 = 100 cores (rounded up for overhead).

  • Processing capacity: 100 cores × 10 requests/sec = 1,000 inferences/sec capacity.
  • Scaling strategy: Horizontal scaling. Adding more instances or containers offers flexibility to meet traffic spikes and simplifies scaling for the projected growth.2. Future capacity planning: -Traffic growth adjustment: 2× traffic over 12 months = 2,000 queries/sec. Required cores = 2,000 ÷ 10 = 200 cores (plus 10-15% overhead for resilience).
  • Load testing: Use synthetic traffic to simulate peak conditions, verify latency targets are met and identify bottlenecks before launch.3. Identify and mitigate supply chain risks Below is an example of how to identify, analyse and mitigate supply chain risks for the Smart Search ML feature.Risk 1: Dependency on a third-party product catalog API-Analysis: The API may experience downtime, latency issues or schema changes, disrupting feature updates and degrading real-time relevance in search results. Likelihood: Medium. Impact: High — could lead to incomplete or outdated product listings.
  • Mitigation:Layer 1 — redundancy: Implement a local cache and fallback mechanism to serve recent results when the API fails.
  • Layer 2 — monitoring: Set up latency and error rate alerts to detect API degradation early.
  • Layer 3 — data contract: Establish an SLA with the vendor specifying uptime and change notification terms.
  • Trade-offs: Increased infrastructure cost for caching and monitoring, but higher availability and user trust.Risk 2: Open-source library vulnerabilities-Analysis: A dependency used in the ranking model (e.g. a pre-trained embedding layer) may contain unpatched security vulnerabilities. Likelihood: Medium. Impact: Moderate to high — could expose the system to injection or privilege escalation risks.
  • Mitigation:Layer 1 — security scanning: Regularly run automated vulnerability scans on all libraries and dependencies.
  • Layer 2 — version control: Use a package manager lock file to ensure consistent versions across environments.
  • Layer 3 — vendor due diligence: Prioritise libraries with active maintenance and transparent update policies.
  • Trade-offs: Requires ongoing monitoring effort but significantly strengthens security posture.Risk 3: Cloud region dependency-Analysis: Relying on a single cloud region may cause outages if that region experiences downtime or resource shortages. Likelihood: Low. Impact: High — could cause total service interruption during peak load.
  • Mitigation:Layer 1 — multi-region deployment: Deploy replicas in at least two regions with automated failover.
  • Layer 2 — load balancing: Distribute requests across regions to maintain consistent latency.
  • Layer 3 — backup strategy: Regularly replicate key data to alternative storage regions.
  • Trade-offs: Higher infrastructure cost but greatly improved fault tolerance and service reliability.

What this example does well​

  • Demonstrates clear, step-by-step reasoning for resource estimation and capacity planning.
  • Provides detailed, layered mitigations for key risks with clear trade-off explanations.
  • Connects planning decisions to both current performance needs and future scalability.

Tips for applying this skill in your role​

  • Add capacity buffers and validate assumptions using staging or synthetic load tests.
  • Document and regularly test mitigation strategies to ensure they remain effective.
  • Use layered, well-justified mitigations that balance cost, reliability and performance.
Skills Solution illustration

Reflection​

Compare your output to the solution example provided. What did you do well? Where could you improve?

  • How would you adapt your capacity plan and risk mitigations if your system’s workload or dependencies changed suddenly?