Self-Hosted Canvas, what are the minimum technical requirementes?

[hnoriega]

Salute, dear community!

We have currently a Self-Hosted Canvas LMS T3-Type hosted in an EC2 AWS server. According to our stress test, the successfull rate is 84% over 500 simultaneous conections. Our server features are 32 GB and 8CPU, unlimited transferences and 640GB storage.

We are planning to have between 500 and 1000 users connected simultaneously, that's why I need your help with the following questions:

1. What are the successfully rate for simultaneous connectios with similar server-features according to your own stress tests? So we can compare and to know if our 84% is normal or not.
2. What are the recomendations for server requirements for 1000 users simultaneously?
3. Do you recomend another kind of setup to increase the sucessfull rate for the connections or to get the 100%?

Thank you very much.

[nicolopez77]

I was asking more or less the same on the google group (is that group supported by Instructure?). Does anybody have some information about RAM/CPU usage per concurrent user? Any experiences to share? 

[nicolopez77]

Hi I'm interested in this topic as well. Any experiences to share? Ideas on how to run a stress test?

Thanks!

[segunola]

Pls this is an important question, why are we not having response from any expert. 

[a4aravind]

Did anyone get any suggestions regarding this....pls share if you've received any support from somewhere. Thanks in advance !

[StevenNgo]

Hi everyone,

Does anyone have an answer for this question?

I am just starting to work on how to self-host Canvas as well, it will be great to hear you all share your experience and insights. @nicolopez77 @a4aravind @segunola @hnoriega 

Thank you in advance.

[Nicolas-Lopez]

Hi Steven, we've hosted Canvas in the past, and still have some active customers, with a configuration of 4 vCPU and 8GB without any issues. They have hundreds of concurrent users with this configuration. Disk size will mostly depend on the course files and user uploads, in any case I can recommend you to move the folder to a separate volume, just in case it gets full. 

All in all, for small implementations the open source version can be a good starting point. For bigger ones the architecture can get a bit more complicated and it makes no sense to use the open source version. It's worth paying the licenses to Instructure and use the infrastructure they provide. We are Instructure's partners, and have experience with both.

Regards,

Nicolas