Adds migration-specific assessment guidance and corrects several configuration defaults in the LMI skill based on testing against real Lambda workloads.

Related

PR awslabs#201 (targeting sbbhimji's branch)

Changes

Formulas for deriving LMI configuration from on-demand metrics (SKILL.md Step 1):

• MaxConcurrency from cpu_total_time and Duration (targeting 50% CPU utilization)
• MemorySize from per-invoke memory utilization × concurrency
• MinExecutionEnvironments from baseline concurrent executions
• Fallback defaults when Lambda Insights is not available
• Instructions for checking/enabling Lambda Insights

Corrections:

• Default memory-to-vCPU ratio: 4:1 → 2:1 (matches public docs, CDK defaults, and our testing)
• Weighted alias traffic shifting: noted as not currently supported for LMI functions
• Best Practices updated to align with corrected defaults

Testing phase guidance (references/troubleshooting.md):

• Added "Testing Phase: Monitor and Adjust" section with metric-specific remediation steps for duration regression, CPUThrottles, MemoryThrottles, DiskThrottles, ConcurrencyThrottles, and low CPU utilization

Questions for reviewers:

• Is there a reason 4:1 was preferred as the starting ratio? Our typical recommendation is 2:1 (the documented default) which we had been thinking works well for most workloads, but curious if there's data suggesting otherwise.

Acknowledgment

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of the project license.