Using Correlation Analysis to Monitor DBMS Performance

Postgres Pro (enterprise certified) 15.7.2 on x86_64-pc-linux-gnu, compiled by gcc (GCC) 11.4.1 20230605 (Red Soft 11.4.0-1), 64-bit

CPU

processor       : 0
vendor_id       : GenuineIntel
cpu family      : 6
model           : 85
model name      : Intel Xeon Processor (Skylake, IBRS, no TSX)
…
processor       : 5
vendor_id       : GenuineIntel
cpu family      : 6
model           : 85
model name      : Intel Xeon Processor (Skylake, IBRS, no TSX)

RAM

free -g

               total  
Mem:              31  

Experiment results

Load on the DBMS

Fig.1. Total number of sessions and number of active sessions during the test

Fig.1. Total number of sessions and number of active sessions during the test

DBMS performance

Fig.2. Graph of changes in DBMS performance during a stress test

Fig.2. Graph of changes in DBMS performance during a stress test

Correlation between DBMS performance and the number of sessions in the waiting state

Fig.2. Graph of correlation changes during a stress test

Fig.2. Graph of correlation changes during a stress test

Results and conclusions

  1. An indicator of DBMS performance degradation can be the correlation value between the DBMS performance and the number of sessions in the waiting state. Increased expectations and decreased DBMS performance indicate the presence of problems and the need to create an incident to analyze and resolve the situation.

  2. The priority of an incident is determined by the value of the correlation coefficient:

    -0.7 < Rxy < -0.5 : Low priority

    Rxy < -0.7 : High priority

Similar Posts

Leave a Reply

Your email address will not be published. Required fields are marked *