According to the exam guide, one of the objectives is to "describe the use cases and considerations for deferring sharing calculations"1. This implies that option A is the correct way to load large amounts of data into Salesforce without affecting performance and data integrity. Deferring sharing calculations allows the data to be loaded first and then the sharing rules to be applied later2. Option B is not correct because workflows are not recommended for calculating summary values, as they can cause performance issues and data skew3. Option C is not correct because triggers are also not recommended for calculating summary values, as they can cause governor limit errors and data inconsistency. Option D is not correct because external IDs are used to link records from different systems, not to improve data loading performance.