Price Anomalies

For most of the resources inside the tenant, prices of individual resources are almost constant, they do not change dramatically from one day to another. Sometimes this expectation is violated when there is unproportional usage. We have developed monitoring for such deviations and report them as price anomalies.

Detecting such anomaly early and remediating can present significant savings.

Origin of price anomalies

Examples that can cause price anomalies: - Resource group that was existing for several months with relatively steady price changes, suddenly grows by 10% or more every day - Database running with 2 VCores for some time is suddenly changed to 20 VCores - 100 GB Standard SSD disk was resized to 4 TB Premium disk

All those and more anomalies would be detected. They share one thing in common, that is they need to have some history of price and suddenly start to change rapidly.

How detection works

In the image below, you can see graphical representation of how we search for price anomalies. We use something called linear regression channel. In the picture, price changes that would be detected and reported as price anomalies are circled in red color.

Example: - Price of resource was between 9 to 10 USD for months - This resource price increased to 20 USD in two days

Now based on these facts: - If the price comes back to the previous range, it won’t be reported as price anomaly - If the price stays at its elevated price of 20 USD or even keeps increasing, it will be reported as price anomaly