Repository Compute Tier (v20.x)

  • 13 July 2023
  • 2 replies

Userlevel 3
Badge +1

What would be the recommended compute tier for a project's repository database? (TX v20.x) 

I believe the current default on Azure is General Purpose 2vCores, however looking at the utilization this seems to be excessive. 

Does anyone have experience with running the repository on a Standard DTU database? 



Best answer by wouter.goslinga 13 July 2023, 09:15

View original

2 replies

Userlevel 2
Badge +1

Hi @sierd.zuiderveld , I commonly use a Standard 20 DTU database for the repository. This works fine, even for larger projects with several thousands of versions:



Userlevel 6
Badge +7


while 20 DTU is a good starting point (until the DTU model disappears) I disagree with: "This works fine, even for larger projects with several thousands of versions:” . There are three parameters that define repository responsiveness:

  • number of objects in project (the export size of the xml is a proxy for this)
  • number of changes in repository
  • amount of execution logging held in the repository database

You will generally notice slowness creeping in as these parameters grow. But really that isn't something to worry about as the whole point of PaaS databases is that you can scale them as required. However, if you know you are going to need the resource for a year or longer you can get substantial cost reductions from reserved capacity. Reserved Capacity is not available in the DTU model.