An application uses AWS Lambda and ECS on Fargate and stores data in an Aurora MySQL database. The workload is write-heavy and traffic is highly spiky with long idle periods and sudden increases. The current memory-optimized DB instance cannot handle variable load. Which is the MOST cost-effective design to scale for these traffic patterns?

Aurora Serverless v1 is ideal for spiky, unpredictable workloads with long idle periods because it automatically scales compute capacity up and down, including scaling to zero, and you only pay for the capacity consumed. This makes it highly cost-effective compared to provisioned instances that you pay for even when idle. Compute Savings Plans further reduce the cost of Aurora Serverless v1 by committing to a consistent amount of compute usage. Adding read replicas (A) helps with read-heavy workloads, but the problem states the workload is write-heavy. Multiple writer instances (B) are not a standard Aurora feature; Aurora typically has one writer. Aurora Global Database (C) is for disaster recovery and low-latency global reads, not primarily for scaling a single region's spiky write workload. Instance Savings Plans and RDS Reserved Instances (A, B, C) are for predictable, steady-state workloads, not spiky ones, and would lead to under or over-provisioning and higher costs.