“MsDyn365FO development benchmark” of some different VM sizes
This is the fifth post of a series which describes how to deploy the downloadable VHD for Microsoft Dynamics 365 Finance and Operations as an Azure VM. If you want to read the previous post, please use the following navigation:
Before we wanted to rollout the new Azure hosted VMs to a broader audience of developers, we needed to measure some performance metrics which will affect the duration of the development process. With this small “D365FO-Benchmark” we wanted to compare following development VM setups:
– On-Prem hosted VM (8 vcpus 2.6 GHZ / 32 GB memory / SolidFire Storage)
– Azure hosted B4ms (4 vcpus 2.3 GHZ / 16 GB memory / Standard SSD)
– Azure hosted B4ms (4 vcpus 2.3 GHZ / 16 GB memory / Premium SSD)
– Azure hosted B8ms (8 vcpus 2.3 GHZ, 32 GB memory / Standard SSD)
– Azure hosted A8m (8 vcpus 2.3 GHZ, 64 GB memory / Standard SSD)
The first metric we wanted to compare is the time needed to compile several packages. Because we are currently on version 8.0.4 with almost all X++ hotfixes included we still have to compile a lot of packages. We decided to compare the compile time of only I few packages, to do not overcomplicate things. The compile time is written to a log file from the compiler and is not manually stopped by using a stopwatch.
The second metric we wanted to measure is the time needed to execute some bigger end to end (process) unit tests. Because the creation and execution of unit tests is a major part of our development process we need fast execution times in this area. I will not go into detail what those unit tests are doing in details, because we only need the figures for the comparison of the VM types.
Conclusion of the “MsDyn365FO development benchmark”: We will go for B4ms VM size with Standard SSD storage because Premium SSD do not have a big effect and also a bigger VM size only has an effect on big packages like ApplicationSuite. What we can definitely see is that no Azure VM achieves the performance of our On-Prem setup. This is most likely due to that fact that we use higher CPU clock rates in our On-Prem hosts and VMs.
By the way: I can´t tell you why A8m is significantly slower than B8ms!