Estimate daily unique users from billions of access-log rows.
A trend estimate is acceptable if its error is measured and communicated.
Which approach is most appropriate?
Use an approximate distinct aggregate for a tolerant use case, and keep billing or audit counts on an exact path.
Detailed explanation
Exact counting can consume more resources at this scale than the trend requires.
Exact counting can consume more resources at this scale than the trend requires.
An estimate can reduce memory, scan work, and runtime when exactness is not required.
An estimate can reduce memory, scan work, and runtime when exactness is not required.
AWS API actors are not the same population as business-log users.
AWS API actors are not the same population as business-log users.
Retention rules do not safely deduplicate source events.
Retention rules do not safely deduplicate source events.
Check official documentation
Compare the following points with the official documentation.
Amazon Athena approximate functions、approx_distinct、COUNT DISTINCTの公式説明を確認する。
What to confirm
厳密集計と近似集計のコスト・精度のトレードオフを説明できる。
Key points
- approx_distinct
- Approximation
- Error bound
- Cost
Notes
- Environment: AWS公式DEA-C01試験ガイドと各サービスの公式ドキュメントの確認
- Confirm current AWS service behavior in the official DEA-C01 documentation.
Foundation review
Use-case separation
Trends can tolerate approximation while billing and audit values may require exact counting.
Validation
Compare samples or exact calculations to quantify error before publishing the estimate.