In Oracle 10g was introduced the "Intelligent Self-Management Infrastructure" [1].
Basically the idea is to collect the most informations with the minor impact on the system. Based on Heisenberg's Uncertainty Principle, we know that "the observer affects the observed" [2]. And this is true also for Oracle.
One component of the "Intelligent Self-Management Infrastructure" is ASH. In this post I would like to show you a special point of view base on an idea of Kyle Hailey [3].
We know that ASH is based on sampling made each second. This mean that there are the gaps between snapshots. Really we want analyze only long sessions (only "this long sessions" impact on the performance). So, if a session takes a lot of minutes before the end, you are sure to observe the work, even if the observation is made every second. Is like you see a movie, one frame each second.
Because a picture is worth a thousand words, I show you this picture from [3].
It shows how is possible to understand what is happening also if we don't know the story in real time.
And the wonderful thing is that the sampling is like the real time. In order to show that this last statement is true I used a query [4] which output is this chart. I created it using Microsoft excel.
Really, to build this chart, I delete some values. I'm investigating the root cause of those anomalies.
- Update 31/Mar/2016
I modified the chart, because I deleted one point from the graph. I'm writing a post on anomaly I observed.
- References
[2] http://www.oracle.com/technetwork/database/manageability/twp-40169-134162.pdf
[3] https://sites.google.com/site/embtdbo/wait-event-documentation/ash---active-session-history
http://blog.orapub.com/20150812/Which-Is-Better-Time-Model-Or-ASH-Data.html
[4] http://www.evernote.com/l/ABzQFwwMxo5HyKs8pmpIcyVjmjH7vreMvVo/
Nessun commento:
Posta un commento