Is the Jupyter Enterprise Gateway still the go to solution for distributing ML work loads to dedicated clusters?

Is the Jupyter Enterprise Gateway still the go to solution for distributing ML work loads to dedicated clusters?

How widely used is this project and is it actively maintained?

Is there a roadmap for future enhancements?

Hi @mcberma - thanks for the post.

Your wording - go to solution - is flattering, but I would view EG as a solution.

I’m not very aware of other solutions but I believe EG is unique in that operates on a kernel granularity level. Other solutions, like Hub deploying Notebook servers on Kubernetes or Kernel Gateway running in the cloud, still create their kernels local to the server. As a result, you can’t take advantage of specific resources on a per-kernel basis and only for the kernel’s lifecycle.

There are some fairly large companies deploying EG although I hesitate to share publically. One customer, PayPal however, wrote a blog post mentioning EG. They also gave a talk at one of the conferences in San Francisco in which EG is referenced, but I’m not able to find that particular video.

The project is maintained by a few individuals (and we could certainly use more help) and we do our best to be as responsive as possible, although there’s only so much time in a day.

One area where we should spend more time is in maintaining our roadmap, although one does exist.

Please feel free to swing by the repo and ask questions there as well.

1 Like

Kevin:

Are you available for a conversation with some members from our team, I will be glad to set up the meeting logistics if you are available and can provide me with dates and times (in the appropriate timezone) that work best for you.

Regards,

Mark Berman
mark.berman@usbank.com
952.292.4882

Sounds good Mark, I’ll send you an email and we’ll go from there.