Typically I would consider the 5,000 Relays the MAX number of Relays the master can expect to maintain IF all those relays are on the WAN and not using NAT. If NAT or Relay in Cloud using ISP public IP address then the Master would need to maintain a TUNNEL in order to be able to start communication traffic vs. having the relay wait until it "checks in". The number would be reduced dramatically for each Relay under NAT.
This customer's Master should have a very good I/O for DATA Storage and fiber to the DB Server. Not good if Postgres installed on Master Server as I/O will go thru the roof. Will each relay be supporting a lot of end points? By a lot, I mean over 1,000 or will they be a parent to very few devices like 10-30 devices. This too will have an impact on the formula.
I would contact Serena and provide all the details so R&D can weigh in on this customer's requirements. You may need to consider a Super Master setup where the load will be split between two master servers and each master would report into the super master that acts like a consolidator of data. Best for Reporting where the heavy lifting is done at each master server.
Another component to consider, will this customer be using all modules such as Patch and Software Deployments? This too will factor into how robust this will be along with reliability of Data Reads and Writes while deploying all the patches, ect..
Hope my response helped but not gospel... Get Serena involved good buddy so it is all configured with the blessing of BMC.
Thank You Steve, your answer is very useful.
On my customer environment each relay supports until 10 devices, the DB is separated and each server has a solid HDD