ODQA, simultaneous users as opposed to Hardware capacity and speed?

DeepPavlov Library Models
1

I understand that in order to run DeepPavlov with a chatbot it is wise to have a permanent online server. Right now I have about 80000 textfiles to search through for my ODQA.

Now I wonder with a certain system how many people can then use the DeepPavlov API simultaneously while maintaining performance? I know it will depend on computer performance, but do you have some kind of graph/chart or other information showing the cpu/gpu power as opposed to number of simultaneous users supported and/or time, so I have an idea as to how it all scales to each other?

Right now I am using a Intel® Xeon® Platinum 8124M CPU @ 3.00GHz (it has 18 cores)
with 20 GB’s of RAM

Thank you.

2

Hi!

I am not aware of any stress tests results for ODQA model. If you could perform such tests and share results we would be grateful.

3

Hi Yura,

But 80000 files for ODQA should be possible right? With a relatively fast computer do you have any indication of how fast that would be. Or speed as opposed to words? You guys must have tested with some type of large amount right?

Kind Regards,

Kostis