GRATS: Multi-Tier Data Storage Allocation for Smart Cities

Municipal data storage absorbs multi-million-dollar capital budgets over decade-long horizons, yet tier allocation happens under documented institutional opacity. We introduce GRATS, a tri-objective framework jointly minimising monthly cost, volume-weighted latency, and operational plus amortised embodied CO2, with data residency as a hard constraint. The problem is cast in the Environments-Classes, Agents, Roles, Groups, and Objects (E-CARGO) lineage as a multi-objective Group Role Assignment with Constraints (GRA++) instance, and solved exactly via augmented ε-constraint (AUGMECON) on a logarithmic grid in COIN-OR Branch-and-Cut. Post-Pareto Multi-Criteria Decision Analysis (MCDA) via TOPSIS and canonical VIKOR over four Toronto-anchored profiles (CFO/CIO, Service Delivery, Equity, Climate Officer) selects compromise solutions. The reference instance (184 roles, 16 agents, 9.49 PB) returns 474 Pareto-optimal allocations in 430 s, with profile picks achieving up to 2.85× cost reduction and 6.21× lower carbon than a lifecycle-on-premises baseline. The formulation generalises to any municipal data ecosystem with comparable governance constraints.