Problem Statement
The bank’s multi-region data center estate consisted of legacy platforms, bespoke middleware stacks, and monolithic applications scheduled for phased modernization. Technology leadership needed to determine the optimal blend of platform rehosting, application refactoring, hardware refresh, and data center consolidation to hit aggressive CO₂ reduction targets. However, fragmented inventories, static carbon accounting spreadsheets, and siloed planning processes made it impossible to quantify how architectural choices (e.g., retiring specific facilities, changing platform mix, adjusting refresh cadence) would impact emissions within budget and time constraints.
Solution
We designed and deployed a sustainability digital twin of the bank’s data center environment, built on graph-network modelling to represent dependencies across facility, hardware, platform, operating system, and application layers. The twin ingested telemetry and asset metadata to simulate how different modernization pathways would influence carbon intensity, capex/opex, and delivery timelines.
Core design elements
- Holistic asset graph: Unified CMDB records, telemetry feeds (utilization, PUE, energy mix), and application portfolios into a canonical graph with lineage across facility → hardware → platform → service → application workloads.
- Scenario engine: Enabled parameterised “what-if” simulations (e.g., decommissioning two legacy data centers, migrating 60% of workloads to a hybrid cloud platform, sequencing OS upgrades before application refactors). Each scenario produced 36-month emission trajectories, budget utilization, and marginal abatement cost analyses.
- Constraint-aware optimisation: Incorporated budget envelopes, regulatory constraints, and capacity limits to surface feasible pathways that met the bank’s emission reduction targets within specified investment windows.
- Stakeholder interface: Delivered interactive dashboards for sustainability, infrastructure, and application owners to evaluate trade-offs, compare pathways, and align on execution sequencing.
Execution approach
- Iterative calibration cycles validated model outputs against actual energy consumption and emissions data, achieving ±6% variance by the third sprint.
- Governance mechanisms captured scenario assumptions, decision approvals, and post-implementation telemetry to continuously refine the twin.
Outcomes
The digital twin grounded sustainability planning in observed telemetry rather than spreadsheets, predicting emission pathways with 94% alignment to the first quarter of actuals. Scenario experimentation compressed decision cycles from eight weeks to roughly two and a half, giving executives time to evaluate trade-offs before budget windows closed. By sequencing modernization waves through the twin, the bank avoided $11.3M in low-yield capital spend and codified the model into quarterly steering governance, turning sustainability targets into an executable, data-backed roadmap.