Showing results for 
Search instead for 
Did you mean: 

CMS 1000 VM resource allocation question

Hi all


     Cisco document says that CMS1000 support up tp 96HD calls. My client purchased a CMS1000

and a SMP+ license for CMS. How should I edit the CMS vm  (vCPU, vRAm...) which was preloaded on CMS1000 in order to support 96HD calls? I can't find any document talking about this.




the CMS1k has 72 logical cores. Leaving 2 for ESXi will result in 70 cores.
Not sure if there is any practical difference between 2:35 or 1:70. CMS will see 70 cores either way.
Jaime Valencia
Hall of Fame Cisco Employee

That wasn't even there, good to know they added some info.

You'd really need to discuss with the BU or a TME the reason for that, which is different from what they send in the pre-configured OVA.

From a technical perspective, I agree with Zoltan, Only certain OSs would have/had specific requirements as to how they required virtual sockets Vs. cores per socket. But ultimately, the performance would be the same as it's really the hypervisor who handles the scheduling.



if this helps, please rate

Hello Lior,


I agree that isn't the clearest, but if you take a look at Appendix C.1.1 Sizing your VM - Call Bridge VM - Running a single VM it mentions that you should configure mirror the underlying hardware.


Running a single VM When running a single virtual machine on a host, one physical core per host must be left unused by apps for ESXi scheduler. With a single VM, it is possible to use hyper- threading to increase the available capacity. In this case the number of available vCPUs is double the number of physical cores in use. So a two socket system which has 20 physical cores will have 19 available to the application. With hyper- threading enabled, 38vCPUs can be used, which should be allocated to the Meeting Server VM, and the other 2 left unused. If an option is available to choose both number of sockets and number of cores per socket, then these should mirror the underlying hardware.


Even in your excerpt it does still say that "On a dual processor host with hyperthreading, set Number of Cores per Socket to the number of logical cores minus 2". Which does suggest that it's still per core. Either way I do agree bullet a. should be clearer.


You definitely could run into a performance decrease if the VM believes it's using 1 processor with 70 cores vs 2 processors with 35 cores each due to how CPU cache works. If the VM isn't aware of the number of cores it may not keep the multi-threaded applications on the same CPU which could decrease the performance for certain tasks. However I can't say for sure the exact performance difference you'd see because there are a lot of factors that could affect how impacting it would be to have a process split between two physical CPUs.

Ok thanks guys but I'm still not sure what to choose
@Jaime Valencia, what is BU / TME?

Hello Lior,


You want to do 2:35 not 1:70 because the underlying hardware is the CMS 1000 uses 2 sockets each with an 18 core (36 thread) processor. We'll open a defect for page 12 to have the Business Unit (BU) for CMS to clarify that bullet that you pointed out because it's could easily cause confusion as it did here.

Wonderful, good answer for me at this point.



Unfortunately, the documentation is still not clear as there are conflicting instructions. The latest doc for 2.6 is even worst, the chapter for OVA deployment was removed and there are broken links in the PDF:


I deploy CMS 2.6.1 on a MM 410v and I struggle to know what are the right CPU parameters.


There are 2 CPU with 12 physical cores on each, and hyperthreading is enabled = 48 logical cores

According to this post, I would say:

2 sockets with 23 cores per socket = 46 logical cores, leaving 2 logical cores for the Scheduler

Regarding vRAM, there are 64GB in the server, I would assign 58GB to the VM.


Please advise.

Thank you very much.


We have CMS cluster 2.8.1 deployed using OVA (


This has 16vCPU and 16 GB ram. Can we increase its vm parameters to increase more than 10 Full HD calls.


Any impact if we change this default OVA parameters or is there any procedure to do changes.





The same issue is happening still in 2020.  CMS 1000 shipped with 8 vCPUs and 16GB of ram.  This baffles the customer and me as well.

Please open a case with TAC, so we can investigate why this happened with your system.

Hi  Zoltan


     Yes. All licenses were applied. Including Vsphere 6 combined key for 2 CPU and CMS activation and call license keys. So far, I have deployed 3  CMS1000 for my clients. All CMS1000 come with preloaded VM with 8vCPU, 100GB harddisk and 16GB RAM.  That's why I am so confused with the VM spec. Thanks!

Unfortunately my box came with preconfigured Telepresence server. when I create CMS OVA it does not let me configure more than 32 vCPU of 1 socket or 16 vCPU of 2 socket.


I right clicked on VM of CMS and click on Upgrade VM Compatibility and my VM machine upgraded from 8 to 11

Sebastien Lemire

I want to chime in on this. We recently got delivery of 8x CMS1K-M5 servers.

I actually had 4x of them with the above listed specs of 8vCPUs but to my surprise, the other 4x had 36vCPU and 128GB of VRAM. The new servers actually include 128GB of RAM. Boggles the mind.


Also I've had this happen last year and in this batch, with the pre-configured VM, no matter what I do, I can't get SSH to work, I get a connection refused. Only way I figured to fix issue is to wipe the VM and re-load fresh from OVA. While I'm at it, I fix the VM specs..

Hi @Sebastien Lemire, completing your recent findings here, and interested to know how you solved it.


Same issue for me, a freshly received CMS1000 M5, still preloaded with OVA 8vCPU and 16GB RAM...

Preloaded with ESXi 7.0 as well.


2 CPU, 18 physical cores each = 36 physical cores

We assign 36-1=35 physical cores to the VM. With HT means 70 logical cores. 2 sockets with 35 logical cores each. OK.

I upgraded the vm version to be able to edit the settings with the 2x35 cores. OK.


For the RAM, 58GB was recommended in this post with lots of valuable experts, but it becomes even crazier now.

On ESXi 7.0, the datastore comes with only 150GB disk space. 100GB for the CMS vmdisk, 50GB left. If you setup more than 48GB of RAM for the VM, ESXi just refuses to start the machine with a "MonitorLoop" error message, because the swap file has not enough space on the disk (needs ~60GB for the 58GB RAM)...


Should we stick to what is written in the installation guide? "1GB RAM for each underlying physical CPU core should be allocated to the VM with a minimum allocation of 4GB of RAM. For the system above, the VM should be configured with 19GB corresponding to the 19 physical CPU cores in use." --> we would only need 35GB?


This page doesn't help: , stating that you need 1GB for each vCPU = 70GB??


I'm very sad to see that after many years, the documentation is still not clearly mentioning what are the static requirements for the VM settings on CMS1000. These are sold as standard appliances, with tested and predefined specs, there should be written black and white how to setup the VM specs...


All feedbacks are welcomed. Thanks!

I've reached out to developers to investigate the issue, and they're already discussing a possible fix. In the meantime, could you please open a case with TAC, so we can work with you to resolve the problem!

Recognize Your Peers
Content for Community-Ad