CODEWITHSUNDEEP
amazon-web-servicesamazon-dynamodbamazon-cloudwatch
DaveR
DaveR

Reputation: 2358

How does dynamoDB calculate read events over time?

I have read the docs about dynamoDB capacity but I do not understand how they calculate their metrics. If you look at the image, how is it possible that I have short bursts of throttled read events on the left when summing them up over 1 minute but just a constant of 1 when averaging them over 1 minute (on the right).

Does it mean that most of the throttled events are being absorbed by the burst capacity and/or is it a sign a should increase the provisioned capacity?

Sum and average over 1 minute

Upvotes: 0

Views: 73

Answers (1)

Mike Dinescu
Mike Dinescu

Reputation: 55720

The way these metrics work is that DynamoDB will emit a value of 0 or 1 indicating whether throttling occurred or not for each request.

CloudWatch performs aggregation on these metrics and produces statistics that you can then examine through the CloudWatch Metrics tool.

In your case, for throttling, AVG is not a very helpful statistic (because the underlying metric is discrete). Instead, you should look at the SUM statistic. That will give you the total number of requests in each aggregation interval that had throttling. In other words, if you are looking at a period of 1 minute, then you would get throttled requests / minute but if you chose 5 minute aggregation, then you'd be looking at # of thottled requests per 5 minutes

Upvotes: 1

Related Questions