Domain 3 · Optimize Usage and Cost
Cheaper architectures, same performance.
Workload Optimization is the architectural sibling to rightsizing — when a different shape, tier, or service would do the same job for less. CloudMonitor flags those patterns with dollar deltas and evidence per workload, so the next ARB doesn't argue from gut feel.
The problem
Arch reviews that skip cost.
Arch reviews skip cost.
Reliability, security, scalability — all in the template. Cost is "we'll check after launch", and after launch is always too late.
Patterns chosen for familiarity.
The team picks the service they shipped last quarter. The cheaper-and-equivalent alternative never gets a hearing.
Migration ROI in slides.
"Move to serverless saves 40%" lives in a deck nobody updates. The actual delta after migration never gets measured.
How CloudMonitor answers
Patterns with prices, evidence per workload.
Pattern library with deltas.
VM-to-App-Service, IaaS SQL to Managed Instance, ADF to Synapse pipelines, container to Functions — each pattern carries a dollar delta and a fit-test.
Performance-preserving.
Recommendations are gated on observed latency and throughput, not just cost. The replacement has to hold the SLA.
Evidence per workload.
One page per recommendation: current shape, target shape, six-month projection, telemetry that supports the call. The ARB can sign off on it.
Tracked after migration.
Approved migrations are tracked against the projection. The actual delta — not the slide delta — is what the practice reports.
Outcomes
Architecture changes the ARB will buy.
Δ$
Dollar delta per pattern
SLA-gated
Performance preserved
Tracked
Realised vs projected savings
See the pattern library against seeded workloads.
Walk a recommendation, see the evidence page, export the ARB write-up.