President & Founder, OraPub Inc. Oracle Ace Director
OraPub, Inc. USA
Imagine being able to prove to yourself and others a performance problem is not a database issue. There are a few ways to infer the problem resides elsewhere, but it’s difficult for people to believe you. However, by creatively using ASH data we can prove the problem does not reside in the database. For example, suppose a batch process is “taking too long” yet you can not find evidence of the process in an AWR report, analyzing AWR data or even standard ASH analysis. The combination of enabling ASH to temporarily collect “non active” sessions, step-by-step watching an Oracle session while paying close attention to the “SQL*Net message from client” event, we can prove a performance issue resides outside of the database! In this presentation, I will demonstrate exactly how to do this. I’ll start with a quick ASH introduction, then dive into Oracle’s time model, what to look for and how to look for it and finally how to interpret the results. Join me as I show you how to prove the database is not the problem!
Oracle performance issues typically fall into two categories. Either “I’ve seen this before and it’s bad!” or “I’ve never seen this before. We better check it out!” The good news is, a trained analyst with many years of experience can quickly do an AWR or ASH analysis. The bad news is, this manual approach DOES NOT SCALE! Even an expert can’t comfortably monitor hundreds or thousands of databases. And, our rule based systems are relatively simplistic, because they can’t capture the complexity and diversity of activity in production Oracle systems. One solution for this unscalable monitoring and analysis problem is to use machine learning. So, in this webinar I’m going to introduce you to the world of applied Machine Learning from an Oracle Professional (DBA/Developer/Manager) perspective. This includes understanding what ML is, why use it and why now. But the best part is, I will demonstrate how to create an automated anomalous performance detection system! I’ll be using industry standard Python with its ML libraries and Jupyter Notebooks. You will be able to download and do everything I do in this webinar! If you have ever wondered how ML can be applied in an IT environment, you don’t want to miss this webinar.
Suppose you’re doing an AWR or ASH based performance analysis and said, “I’ve NEVER seen this before! We’d better check it out.” What you just identified was an anomalous performance event that warrants your attention. With the help of machine learning and the usual AWR and ASH data, we can confidently detect an anomalous performance event and alert an Oracle performance specialist in under a second! How to do this step-by-step, is what this presentation is all about. Join me as I build a “Performance Anomaly Detector” using 100% free industry standard tools, such as Python with machine learning libraries and the “Always Free” Oracle Autonomous Database with a Jupyter notebook. All presentation slides, scripts, notebook and a recording of the demonstration is available!
Most Oracle performance alerting is based on simple rules. For sure, some statistics are used, but still it’s based on simple rules. With machine learning (ML) we easily go beyond simple rules, because ML algorithms are built to recognize patterns in data. The pattern recognition ability of ML goes far beyond what even a highly trained performance expert could ever hope to achieve. However, most monitoring and alerting systems do not truly use ML, many IT organizations are not ready to embrace ML and many DBAs are not cross-trained in ML. A novel solution to this problem is to use ML to create a list of rules that define “poor performance” based on your real system and real users. These rules may not even make sense to us, but they will take your alerting to a higher level because they are based on your data and ML algorithms. Join me for a fascinating presentation about how to use the results of ML as inputs into your existing monitoring and alerting systems.