Comparing cloud vs on premise
We are researching moving our application to the cloud. Primarily we are looking at our batch tier and how it performs in the cloud vs on premise. Our batch runs multiple jobs in parallell, some run for hours. Each is run in its own JVM, and is represented in a separate purpath.
We have Dynatrace in the cloud and on premise. We run the same batch jobs with the same data in each location. Just glancing at the first testing results looks like i/o methods take longer in the cloud. For example java.io.unixFileSystem methods.
I am looking for a way to compare the set of batch jobs on premise vs in the Cloud. For example is there a way to compare the method breakdown for the set of batch jobs to understand the method hotspots in each location?
Hi Kathy
I would build a dashboard that has the API and Method Dashlet on it and then set the source to a session that contains a PurePath for onPremise and the Comparison Source to the session containing a "cloud" PurePath.
The challenge in your case will be that most of these methods that you see, e.g: unixFileSystem are probably picked up by our Auto Sensors - correct? They show up with a grey icon in the PurePath Tree?
Right now there is no good comparison of Auto Sensor Data. But - if you start with the API Breakdown you will see which API shows the biggest difference. The Method Dashlet will give you an indication on which "instrumented" methods have a difference. Then I would open two Method Hotspot dashlet side-by-side. The Method Hotspot doesnt provide a built-in comparison feature like API or Method, but - you can have them side-by-side and set a different data source for each. Then you can compare and see which method hotspots are different -> it is a visual comparison but it should work well - especially because your purepaths are running that long and the hotspots + differences should be easy to spot
hope this helps
Yes most methods are from auto-sensors.
There are hundreds of batch jobs (and purepaths) in a batch run. I was hoping for a way to characterize the entire batch run for all purepaths. I.E. java.io methods are 20% slower i the cloud.
Is your recommendation analyzing one purepath at a time?
Hi Kathy,
The API Breakdown comparison as outlined by Andreas above should help you achieve this at the larger scale. If you have particular packages you're concerned about you can define an API for these methods in the System Profile for a more tailored comparison via this dashlet (System Profile - APIs), otherwise we will model the data into known/detected APIs.
Hope that helps,
Rick B
