The workload is an excellent candidate for preemptible, currently called spot, ECS capacity because it is short-lived, batch-oriented, horizontally scalable, and explicitly tolerant of interruption. The question states that if an instance fails, another instance can reprocess the job. That removes the main architectural objection to using interruptible compute.
Alibaba Cloud describes ECS spot instances as unused ECS capacity offered at substantially reduced prices compared with standard pay-as-you-go instances. The documented recommended scenarios include short-term tasks, data analytics, and applications that are highly fault tolerant and able to tolerate interruption.
Savings Plans and reserved-instance-style commitments can reduce costs for predictable long-term baseline consumption, but this workload runs only at two scheduled times each day. Purchasing committed capacity for a sporadic workload can leave paid capacity economically underutilized. Higher-specification pay-as-you-go instances also do not inherently minimize compute cost.
Using a launch template with preemptible/spot ECS instances allows Auto Scaling to add inexpensive compute when the batch workload requires it, while the application ' s retry mechanism handles interruptions.
Study Guide reference: Core Compute Infrastructure Deep Dive; Scalable Architecture on Alibaba Cloud; cost optimization for elastic compute.
==================================================
