Hello there,
I have data of a number of virtual machines (VMs), which shows how the VMs have been consuming computing resouces (I/O, CPU, RAM etc) for a perion of 30 days. The data shows that for each resources, comsumption is varying alot with several peaks and Troughs.
How do I detect peaks for the different time series data - my goal is to shows that the VMs data peaks simulateneously or not peak at the same time
Regards.
Derdus
One possibility is from the field of speech-recognition. They use DTW (https://en.wikipedia.org/wiki/Dynamic_time_warping) to check similiarity. However, DTW allows comparison on the horizontal and vertical axis and thus the resulting measure is hard to compare.
Btw. do you have labeled time sereis? That is, do your time series have time stamps? You could use anomaly detection and compare then the found anomalies if they have the same timestamp
Hello André Bauer ,
My time series has timestamps. If I pick a point, I can tell the time it happened. Is there like a statsitcal technique or algorithm or a tool that I can use to achive that. I actually have 520 such time serieses each with over 8000 data points.
.....................
Regards
Derudus
What I would do is to first visualize the data. But in this case you have a large volume of data, so I would first extract data that are more salient. In this case it is the peaks.
Once you identify the regions where there are peaks, you can plot charts (across data from all VMs) to visualize the data. That would only give you a basic idea of the landscape. But I think this is important.
To get a statistical answer, you can count the number of simultaneous peaks across VMs. That is, when a peak happens, you can check how many other VMs peak during the same window. So this is similar to results of rolling a dice. I think you can figure out the rest from there.
The most frequent values can be extracted using statistical software, as well as high values.
Haydar Jasim
What do you mean by frequesnt values and what is this software?
- Badges
- Science method