[Question] Does advisor parallelize operations?

[tszumowski]

I was curious advisor runs multiple trials in parallel or is it all sequential? I saw there was a kubernetes configuration so was wondering if it leverages scaling there for parallel optimization.

[tobegit3hub]

Thanks for asking @tszumowski .

It is still under development and it will support parallel optimization with multi-process or Kubernetes. Any contribution is welcome if you are interested in getting involved.

[ibayer]

@tobegit3hub Thanks for releasing advisor!

it will support parallel optimization with multi-process or Kubernetes.

I can you elaborate on this? Are you referring to parallel optimization of the suggestion algorithm? At least for proposal algorithms like random search I would assume that parallelization is rather trivial.

Can advisor be used in a setting with one adviser server and multiple adviser clients without blocking? If not what needs to be changed? I might be interested to contribute this is the effort is reasonable.

[tobegit3hub]

Thanks for your concern and of course you can help.

Now advisor is the HTTP server and can accept parallel requests from multiple clients. All you need to do is run advisor in Kubernetes with our yaml file.

[ibayer]

@tobegit3hub

Now advisor is the HTTP server and can accept parallel requests from multiple clients.

Thanks that's consisted with the small experiment we run. :-)

All you need to do is run advisor in Kubernetes with our yaml file.

Just to be very clear (it confused me initially), it doesn't matter how you run advisor server be it docker run, Kubernetes pip install etc. it will allow you to run multiple trails in parallel.