The Next Wave of AIMay 3, 2023
The Most Recession-Proof Jobs in the Tech IndustryMay 22, 2023
The MASF/SETDS method of detecting changes and anomalies in performance data, its recent implementation and the way to use and interpret the results will be presented with real examples against MangoDB testing data.
Since 1995, a time when the CMG conference published a very influential paper about the MASF method of anomaly detection, this topic has been increasingly more popular in the area of Capacity and Performance management. The modification of the MASF – SETDS method – was introduced in the 2002 CMG best paper and then got implemented in several companies and applications. This included the CMG online class “Perfomalies (Performance Anomaly) Detection”. Most recently this method has been turned into the cloud based serverless API microservice which is available for free via “Perfomalist” web app. The Perfomalist change points detection API was used against MongoDB’s testing data and got acceptable results which were published in the SPEC.org 2022 conference. This paper included an additional post-processing algorithm (XGBoost) to eliminate false positives which is planned to be added to Perfomalist service.
Anfisa Trubina, Data Scientist & Arvid Trubin, Python Developer at Trutech Development, LLC
Both co-authors work on TTD’s perfomalist.com project. Anfisa is the data scientist who has built the main anomaly and change point detection model. Arvid is the Python developer who has written the code to implement the model as a web application in the cloud .
Arvid has prepared the video instructions for CMG on-line class “Performance anomaly detection”
IMPACT 2023 Proceeding Session Video: