Which solution will meet these requirements with the FEWEST possible read capacity units (RCUs)?
Read the table by using eventually consistent reads.
Read the table by using strongly consistent reads.
Read the table by using transactional reads.
Read the table by using strongly consistent PartiQL queries.
Explanations:
Eventually consistent reads provide lower latency and use fewer read capacity units (RCUs) compared to strongly consistent reads. Since the microservice can tolerate stale data, eventually consistent reads are appropriate.
Strongly consistent reads consume more RCUs because they guarantee the most up-to-date data, which is unnecessary in this scenario, given the tolerance for stale data.
Transactional reads provide consistency across multiple items, which is unnecessary for reading a single customer record. They consume more RCUs and are not needed in this case.
Strongly consistent PartiQL queries are similar to strongly consistent reads and would consume more RCUs, which is not required for this scenario where stale data is acceptable.
I value that the answer is:
Read the table by using eventually consistent reads.