Sametime Meetings 11.5 - sizing - performance . nº of users

I had Sametime on premise working fine for 2000 users.

I want to install Sametime Meetings 11.5 in the cloud for bandwidth reason.

Requirements mini from HCL are 16 physical RAM, 500Gb hard Drive, and 8 core.

Question 1:

Should the Sametime meetings be installed on a cloud Bare metal server or normal cloud VM to work properly ?

(I want to rent the machine, with CentOS, + Docker on it)

Question 2:

It is a performance question also. How many meetings does it support at the same time with that specs ?

BBB use to need a Bare metal server.

Hi Dominique, as I know st meetings it's supported on docker for up to 200users. It's for small deployments or poc. If you have to guarantee meeting service for 2k users you should go with Kubernetes, managed by you or as a service.

The 200 users number depends on the size of the VM as well as other activity (how many recordings, etc). He will not reach 200 users with 8 Cores, 16GB which is smaller than the minimum requirements.

Please see the FAQ for details: https://help.hcltechsw.com/sametime/11.5/admin/docker_kubernetes.html

Thanks,

Casey Toole, HCL

Hello Dominique,

we installed our Sametime 11.5 Meeting Server on a small rented vServer (4 GB RAM, 40 GB Harddisk, 2 vCPU) running CentOS 7 and Docker CE.

First tests with 5 Users streaming Video showed no issues (only the "blur background"-function was not really usable, frame-rate dropped significantly, but that also occured on the "meet.jit.si" open-source Version that Sametime Meetings is based on).

Maybe you can check https://community.jitsi.org/t/recommended-server-specs-for-2020/32041/22 for a comparision (should be roughly the same).

So it really depends on the number of users/streams that will connect, the minimum System Requirements are a bit exagerated for a small deployment.

Hope this helps,
Patrick

Hi @Patrick Schneider, which vServer (provider/hoster)?

I'm trying to get it to work on a vServer from Strato, but /dev/snd is missing, thus sametime-meetings-115_jigasi_1 doesn't start.

Hello @Thomas Bahn

it's a CPX21 from Hetzner ( https://www.hetzner.com/cloud ) . We had to upgrade to a 3 vCPU Model because we ran into some CPU-Limitation issues if the number of participants reached 6.

Did you make sure to follow the instructions on https://help.hcltechsw.com/sametime/11.5/admin/configure_virtual_devices.html to create the /dev/snd folder + contents?

Best regards,

Patrick

Thank you fur the answer and for the hint - I have overlooked this point completely.

I couldn't create the (virtual) sound devices.

Seems not to be possible due to the virtualization technology Strato is using (Virtuozzo). Found some hints refering OpenVZ, which is a "sibling" of Virtuozzo as far as I understand.

@dominique perarnaud

Sizinq questions is not something that can be addressed in a forum format

If you require assistance with a full Kubernetes or Docker deployment, please reach out to HCL Services or one of our HCL Business Partners to inquire about professional services.

Please see : https://help.hcl-software.com/sametime/11.5/admin/docker_kubernetes.html

Hi Dominique,

The minimum requirements are 8 Cores, 32GB for the Meeting Server. https://support.hcl-software.com/csm?id=kb_article&sysparm_article=KB0082513

For your first question, you can deploy on a CentOS VM in the cloud. Keep in mind as you are planning your deployment that the Meeting server still needs connectivity to the Sametime Proxy server and MongoDB.

If you are planning on Docker, there are some limitations to be aware of. The Docker style deployment does not autoscale, it is limited to two simultaneous recordings, and the number of users and/or meetings will depend on the size of your VM. Feel free to ask me a question if you don't see an answer here: https://help.hcl-software.com/sametime/11.5/admin/docker_kubernetes.html

You can install Sametime Meetings in a Kubernetes cloud (such as Amazon's Elastic Kubernetes Service, and Google's GKE - Google Kubernetes Engine) . Please note that the cloud vCPUs are not identical in processing capacity to a physical core. HCL will be publishing some case studies on deploying Sametime Meetings on the Amazon EKS Cloud.

For your second question, HCL recommends you contact our technical advisors team for a sizing to determine the resources needed for your deployment. To initiate this process click here to contact HCL.

Thanks,

Casey Toole, HCL

As a follow on from this whats the minimum for a Kubernetes build as in how many servers are required?

We are not using it in anger and the most we would ever have in a single meeting would be 20 people and concurrent meetings wouldn't be a lot, however when doing this with Docker-Compose and the Preview version we had very mixed results and not a great overall impression of it so I thought I would rebuild it in Kubernetes and Helm instead but I am struggling with building this on a single POC box as all the yaml files seem to be set as though there are multiple nodes and servers etc.

It would be interesting to know what is the experience that others are having between a Docker-Compose and a Kubernetes build.

Hi Richard,

For your 20 users, a Docker deployment should be sufficient if you have at least the minimum requirements 8 Cores x 32GB RAM. I recommend you try it with the gold code instead of the beta/preview. If you have trouble with it, you can open a case with our technical support team so we can investigate the issue.

To answer your question about Kubernetes its really a question of what kind of redundancy/autoscaling you require. You can reach out to our HCL Technical Advisors for a sizing estimate. To initiate this process click here to contact HCL.

Thanks!

Casey

Thanks Casey,

Just as a quick follow up if I did want to run this on a single node single server Kubernetes Instance would this be possible? If the answer is no it has been designed from the ground up in a Kubernetes instance to be scaled over several machines then thats the answer and I am trying to do something thats impossible or at least never designed to work in that way.

From the documentation it doesn't seem like thats the case as the kubernetes quick start is al about a single server and there is even the PV for a single node but in doing the install several times now I am not convinced and the only option for a single server instance is Docker-Compose.

Hi Richard,

I added another k8s node to our cluster with the requisite 8 cores/32GB. I edited the helm files with the desired node affinity and taints/tolerations so that everything would deploy to that node only. Seems to work ok, but we haven't use it in earnest yet.

Regards
Heidi

Hi Heidi,

Thanks for this. I have been using the Docker-Compose for single instances and I am just not convinced it performs very well. Everything works but it seems far from flawless and the experience isn't great sso wanted to do a single node Kubernetes deployment to see if that was any better and would be a better option for a customer who can't deploy many servers but still want to use STMeetings.

I will revisit it again and have another look at the helm files to see if I can get it working and try to compare both options.

By default Docker containers will just use as much cpu/memory as is available. However, you can set limits when you start your containers. I honestly haven't looked at the docker-compose for Meetings to see if this is already the case though.

https://docs.docker.com/config/containers/resource_constraints/

Hey Richard,

The Kubernetes Quick Start was not designed to be operational on a single node. It was provided in our documentation as an introduction to Kubernetes only. HCL Sametime Meetings on Kubernetes requires a full Kubernetes environment. We'll be reviewing our statements around this to better clarify.

Thanks,
Trevor