Comparing new D365FO VMs performance
Together with Anton(https://www.dhirubhai.net/in/ahtoh/), we did some research about the latest development VMs performance. Not the comprehensive, but several interesting findings
We tested 10.0.25 and a new 10.0.27 on a different configuration (3 SSD 128GB, 15 HDD 32GB)
“Hello world” test – we created a simple class (in the FleetManagement model) that displays a text and executed it by changing the text string. This probably represents the most common development scenario – you write some code, run it, then change something, run it again, etc…
There was no difference between SSD and HDD configuration, but there was a huge difference between D12v2 and new D8V5 VMs(both Intel and AMD options)
On D12v2 the time between you press start and seeing the result was around 100seconds, for D8V5 - 55-60 seconds.
In 10.0.27 Microsoft introduced a new debug setup, now you can select which modules to load during the debug process. Unfortunately, all setting leads to either slower(by 5-10 seconds) or the same execution time. We also selected 4 modules(AppSuite, AppPlatform, AppCommon) in 10.0.27 and in worked slower(to 5 seconds) than in 10.0.25 with all modules available for debugging.
领英推荐
There was an interesting discussion about HDD and SSD usage. We did several tests related to that:
So it seems even SSDs are slower by IOPS, they are somehow faster during the initial boot time. So if you shutdown your VMs, SSD may be a better choice. However, on my last projects, I try to agree not to use auto-shutdown, even with SSD it may take 15-20 min of waiting time every day until a VM warms up. In this case, there will be no difference, the HDD option is just cheaper.
Conclusion:
Software Engineer at Microsoft
2 年It's all matter of CPU count in a VM. D12 has 4 vCPU, D8v5 has 8 vCPU. That's why D8v5 is faster. It's better to compare performance of more equal systems like D13v2 vs D8v5.
Technical Architect, Dev Lead at Patagonia
2 年This is a great update. We ran into a similar finding three weeks ago, but went with E4ads v5 instead of the D8V5 type you are using. I don't have synthetic benchmarks handy, but I'd venture to guess D8v5 ($280/mo) is slightly faster than E4ads v5 ($191/mo) due to newer architecture. Did you test Standard vs Premium SSDs? We dropped from Premium to Standard (4x512GB); same performance in almost all scenarios (bacpac import, synthetics) but dramatically decreased costs. I've found that the D365 Extension in Visual Studio is the biggest pain point. Compile time improvements help, but general navigation being slow and creating huge pauses is a bigger pain for many on my team. The higher single threaded performance with these new VM types (IPC and turbo) is a huge win. That's partially why we've had no major issues with 4vCore VMs. Have you found anything similar around single threaded performance being the main driver for day to day?
Technical Architect at AgileCadence
2 年Thank for the testing and the post Denis Trunin, it is very helpful
Senior IT Specialist
2 年HI Denis, thanks for your research, you got yourself a follower there :)
MVP | Solution architect | French speaking | Microsoft Dynamics AX & D356FO Technical consultant
2 年I have being using D4V2.