nagios conference 2012 - nicholas scott - advanced data analytics for nagios

Download Nagios Conference 2012 - Nicholas Scott - Advanced Data Analytics For Nagios

If you can't read please download the document

Upload: nagios

Post on 25-May-2015

1.426 views

Category:

Technology


1 download

DESCRIPTION

Nicholas Scott's presentation on advanced analytics Nagios. The presentation was given during the Nagios World Conference North America held Sept 25-28th, 2012 in Saint Paul, MN. For more information on the conference (including photos and videos), visit: http://go.nagios.com/nwcna

TRANSCRIPT

2. Disclaimer Math may occur later. I apologize in advance.2012 2 3. Abstract Introduction Capacity Planning Component Features Different Forecasting Methods When to use RRD Analysis Tool Statistics Pillow Talk 20123 4. Introduction Nagios Data Gathering AttributesSO MUCH DATA (TOO MUCH?)Generally noisy Sources usually not simpleHow many factors are affecting service X on a given host Y?We have data showing X is like this but why?20124 5. Capacity Planning Terminology Residuals Variation that exists after fitting Period A frame of time where a pattern cycles through a complete iteration Example:2012 5 6. Capacity Planning/home/nscott/Documents/NWC Presentations/DataAnalytics/capacityplanning/capacityplanning.mp4 2012 6 7. Capacity Planning Holt-WintersGreat next-step forecasting for complex systems20127 8. Capacity Planning Gets Dicey for anything more, tradeoffs2012 8 9. Capacity Planning Least SquaresBetter for simple trending, obviouslyFinds trend line that minimizes the sum of theresiduals squaredLess computationally expensive than HW2012 9 10. Capacity Planning Good choice for noisy data Possible future mean value 2012 10 11. Capacity Planning Linear Algebra is fun Linear Algebra is grindy Linear Algebra is a great way to really think about algorithms RRD Python abstraction class is available 201211 12. Capacity Planning Quadratic/Cubic Fit Naive Experimental Fits a polynomial of given order to data 2012 12 13. Capacity Planning For quadratic or cubic datasets User decision2012 13 14. RRD Analysis Tool Goals General stats, mean, variance, etc Also do derivatives, multiple order derivatives Bivariate correlation Dependencies: Python >= 2.4 numpy, rrdtool, scipy, matplotlib, mako2012 14 15. RRD Analysis Tool Example running of this thing: ./analyze.py -H localhost -S Current_Load -s201215 16. RRD Analysis Tool Why do you want to smooth your stuff?Noise noise noiseComedy Option: Pretty graphs Mean Stddev Variance2012 16 17. RRD Analysis Tool Derivatives xQuick refresher: y Actual form well use: y t y t1 y t yt 1 = t t t t1 RRD Resolution201217 18. RRD Analysis Tool Uses? Relateable to physics?PositionVelocityAccelerationJerk (seriously) 2012 18 19. RRD Analysis Tool Example, first derivative on CPU Load: analyze.py -H localhost -S Current_Load -d 1201219 20. RRD Analysis Tool Direct use case? Back to bytes/sec 2012 20 21. RRD Analysis Tool Second derivative (acceleration) analyze.py -H localhost -S Root_Partition -d 1,2201221 22. RRD Analysis Tool Bivariate AnalysisCompare two possibly related variablesDefine a relationshipGraph them on the same graphFind Pearsons Correlation Coefficient2012 22 23. RRD Analysis Tool Example: analyze.py -H localhost,localhost -S _HOST_,PING201223 24. RRD Analysis Tool Example: analyze.py -H localhost,localhost -S HTTP,Current_Load201224 25. RRD Analysis Tool Example: analyze.py -H localhost,localhost -S Current_Load,Root_Partition 2012 25