This paper presents FORTE, a framework for optimizing request routing and traffic engineering to balance access latency, electricity cost, and carbon footprint in large-scale Internet applications. The authors show that FORTE can reduce carbon emissions by up to 10% without increasing latency or electricity costs, and by over 25% over three years through optimal datacenter upgrades. They also demonstrate that carbon taxes or credits are impractical for incentivizing carbon reduction by Internet service providers. FORTE uses a flow optimization approach to dynamically adjust traffic routing based on regional electricity costs and carbon footprints. It allows operators to navigate the three-way tradeoff between latency, cost, and carbon emissions. The framework also includes Fast-FORTE, a heuristic that is 20 times faster than FORTE but maintains near-optimal performance. The authors analyze the carbon footprint of datacenters, the variability of electricity generation, and the impact of datacenter upgrades on carbon emissions. They show that datacenters located in regions with cleaner electricity sources can significantly reduce carbon emissions. FORTE is evaluated using Akamai's dataset, and results show that it can reduce carbon emissions by up to 20% while maintaining low latency and electricity costs. The study highlights the importance of considering regional electricity generation and carbon footprints when optimizing datacenter operations.This paper presents FORTE, a framework for optimizing request routing and traffic engineering to balance access latency, electricity cost, and carbon footprint in large-scale Internet applications. The authors show that FORTE can reduce carbon emissions by up to 10% without increasing latency or electricity costs, and by over 25% over three years through optimal datacenter upgrades. They also demonstrate that carbon taxes or credits are impractical for incentivizing carbon reduction by Internet service providers. FORTE uses a flow optimization approach to dynamically adjust traffic routing based on regional electricity costs and carbon footprints. It allows operators to navigate the three-way tradeoff between latency, cost, and carbon emissions. The framework also includes Fast-FORTE, a heuristic that is 20 times faster than FORTE but maintains near-optimal performance. The authors analyze the carbon footprint of datacenters, the variability of electricity generation, and the impact of datacenter upgrades on carbon emissions. They show that datacenters located in regions with cleaner electricity sources can significantly reduce carbon emissions. FORTE is evaluated using Akamai's dataset, and results show that it can reduce carbon emissions by up to 20% while maintaining low latency and electricity costs. The study highlights the importance of considering regional electricity generation and carbon footprints when optimizing datacenter operations.