CODEWITHSUNDEEP
azure-timeseries-insights
vasudha jain
vasudha jain

Reputation: 29

Can we write a custom Aggregation operation to calculate percentiles value in Azure Time Series Insights Gen2

My Time Series Instance looks like this - {"timestamp" : "2021-01-21T08:05:29.575Z", "latency" : "1082", "threadname" : "thread-1"} And I want to calculate 90 percentile latency in every 1 min interval. Default supported aggregate functions in Azure Time Series Insights Gen2 are min(), max(), sum(), avg(), stdev() etc. However there is no aggregate function to calculate percentile over an interval. I wanted to know if there is any way that I can use to calculate percentile value in Azure Time Series.

Upvotes: 0

Views: 190

Answers (1)

Radhakrishnan Srinivasan
Radhakrishnan Srinivasan

Reputation: 98

If you are sure that the given set of samples form a normal distribution, you could probably using the following expression

Percentile = μ+zp×σ
μ-mean
σ-standard deviation 
zp is z-score; zp for 90 percentile is 1.282

TSI has functions to calculate mean and standard deviation

Caution: However this will work only if the distribution is normal distribution. So if you are not sure, its better to wait until TSI supports such function natively. Please use the following feedback form to provide feedback and request new features! https://forms.microsoft.com/Pages/ResponsePage.aspx?id=v4j5cvGGr0GRqy180BHbR2WAy9J-b6NJkUYDF7Ui1zVUN0o2NDNNWFA1TjdITFdKSUZNTFoyVlhYVi4u

Upvotes: 1

Related Questions