Realtime Anomaly Detection on Serverless Monitoring

Doğuşcan Namal, Thundra

Serverless monitoring is different from traditional monitoring solutions, since you can’t access the function’s container.

Wrapping lambda functions is a way to solve that problem. We were able to collect trace, metric, and log stats on AWS Lambda on a production environment by using wrappers on functions.

I would like to briefly present how it’s possible to collect these data and how async monitoring decrease overhead to send that data to the server side. Later, I’ll explain how to process this real-time data stream from firehose with Kinesis Data Analytics, using the moving average algorithm to detect anomaly behaviors on functions.

Required audience experience

Basic understanding of AWS Services is useful but not required.

Objective of the talk

The audience will learn which metrics are important for them to observe their AWS Lambda and how that can be achieved by wrapping functions. Furthermore, they’ll grasp how Amazon Kinesis Data Analytics is helpful to process real-time data from Firehose on anomaly detection.

You can view Doguscan’s presentation below:

Track 2
Location: Euston Date: November 12, 2018 Time: 2:20 pm - 3:05 pm Doğuşcan Namal, Thundra Doğuşcan Namal, Thundra