By using AWS re:Post, you agree to the AWS re:Post Terms of Use

CloudFront Popular Objects numbers are not adding up over different date ranges

0

In the CloudFront console, using the popular objects report, if I select a series of single days, I get different numbers for the Requests column than if I select the same days as a range.

Note that in order to select a single day, you have to confusingly pick two days in the date picker, i.e. selecting 10/16 and 10/17 will show data from midnight on the 16th to midnight on the 17th. (You can verify this by looking at the date stamps in the downloaded file. I think it's a recent change.)

If I select the 16th-17th and the 17th-18th, and add those values together for a particular object, I get a larger number than if I select the 16th-18th. Can anyone explain this?

asked a month ago33 views
1 Answer
0

The discrepancy you're observing in the CloudFront Popular Objects report is likely due to how CloudFront calculates and reports these statistics. CloudFront counts requests for all objects in 10-minute intervals starting at midnight and maintains a running total of the top 150 objects for the next 24 hours. This method can lead to some inconsistencies when comparing single-day reports to multi-day ranges.

For objects near the bottom of the list, there's constant fluctuation as objects move on and off the list. As a result, the totals for these objects are approximations rather than exact counts. The top 50 objects tend to remain more stable on the list, so their totals are generally more reliable.

When you select a multi-day range, CloudFront provides a cumulative view of the most popular objects over that entire period. This may not always match the sum of individual day reports because:

  1. An object that appears in the top 50 for multiple single days might not rank as high when considering the entire range.
  2. The approximations for less popular objects can compound over multiple days.
  3. The running 24-hour totals may not align perfectly with calendar day boundaries.

It's important to note that while these reports offer valuable insights into object popularity, they are not designed to provide perfectly consistent arithmetic when comparing different date ranges. The reports are most useful for identifying trends and the most consistently popular objects rather than for precise request counting across arbitrary time periods.
Sources
View CloudFront popular objects reports - Amazon CloudFront

profile picture
answered a month ago
  • I'm seeing it happen with the most popular object, which is consistently available in the report. Also, the discrepancy is too big to be accounted for by a ten minute buffer around midnight, or similar effect. A concrete example:

    date. # requests 10/16-17 64,753,300 10/17-18 61,157,395 sum: 125,910,695 10/16-18 94,721,374 difference: 31,189,321 % diff: 24%

You are not logged in. Log in to post an answer.

A good answer clearly answers the question and provides constructive feedback and encourages professional growth in the question asker.

Guidelines for Answering Questions