February 2014 Summaries

A customer noticed a discrepancy between network metrics reported by Datadog and AWS CloudWatch, with Datadog's values appearing off by a factor of 2. The issue was resolved by normalizing the timeseries data across different collection intervals and scoping collected data for accurate comparison. It was discovered that system.net.bytes_rcvd had an additional dimension (device) not present in CloudWatch, resulting in differing metric calculations. By adjusting the query to include the same network interface as CloudWatch, comparable values were obtained. This highlights the importance of using tags as query dimensions for comparison when dealing with similar but slightly different values reported from various sources.

The author of the text is investigating why network metrics produced in Datadog are off by a factor of 2 from those seen in the AWS CloudWatch console. The discrepancy was found between the NetworkIn metric in CloudWatch and the system.net.bytes_rcvd metric in Datadog, which reports at a frequency of 15 seconds compared to CloudWatch's 5-minute intervals. To compare these values, the author applied mathematical operations to normalize them to a comparable scale, revealing that the issue lay in the different collection intervals and data points being reported by each platform. The use of tags as query dimensions helped resolve the discrepancy, allowing for more accurate comparison between the two metrics.