What Is a Turnkey Machine Learning Development Server?
The Puget Systems Machine Learning/AI Development Servers are pre-configured, simple to use, web browser accessed machines utilizing JupyterHub, JupyterLab and the Cockpit system-administration interface. Ready to use software stack includes ML frameworks and development tools, TensorFlow, PyTorch, Anaconda Python, etc.. The configuration is easily extensible for other frameworks and environments both on a system wide basis or by individual users. The underlying Linux foundation together with NVIDIA GPU accelerated computing provides a solid base for modern ML/AI working environments and is easily monitored and managed for single or multiple users with the browser based Cockpit interface.
Are These Systems Right For You?
These systems could be right for you if:
- You need a shared remote resource for ML/AI development work for a small team that can self-manage resource sharing.
- You can access the systems from a LAN or VPN. The configuration is not intended for use on public networks. (... but could be with proper firewall protections by your network administrator)
- You are teaching a class or workshops and need to temporally create accounts for several users. Note; GPU resources are limited but for many educational purposes CPU will provide support for many users.
- You want a high performance, nicely configured Machine Learning system for your personal use, but want to work from the comfort of your own preferred OS and device. Note: it is possible to easily add a desktop environment to the "tower" system for use as a conventional Workstation that still provides the advantages of the server configuration.
Please request a consultation, via the form above, if you have questions.
What Is Included?
Why Choose Our Turnkey Development Servers?
Easy setup process so you can get to work faster
- Use Cockpit to add users, login through JupyterHub or MS VScode and start working
Remote Access via any modern web browser on any OS
- Fully utilize the system from Windows, MacOS, Linux, or any device with a web browser
- Connect from a PC, Laptop, ChromeBook, even a Tablet or Cell Phone
Multi-User or Single-User
- One system allows easy access for your entire team, class, or workshop
- Simplified single-user/power-user experience
High performance hardware reduces development cycle iteration time
- GPU accelerated computing and many-core CPU's
Ready to use with state of the art Machine Learning frameworks pre-installed
- Easily add/remove customized environments to JupyterLab for all users or by individual users
Simplified system-administration means no specialized Linux knowledge required
- Browser-based administration interface for system maintenance and user management
- Easily add and remove users, apply updates, and monitor usage
- Terminal access if needed from Cockpit, JupyterLab, VScode and SSH
Task oriented, "how-to" documentation included to keep you productive
- Administrator and user guides to help with customization
Machine Learning Server Details and FAQ
Q: Do you have Docker container support in the configuration?
A: Not yet but coming soon. Several container solutions have been tested. The configuration is very "user oriented". We want a container system that maintains the same user experience as running locally installed applications. NVIDIA's ENROOT system is the likely candidate. It allows docker containers to be used in a familiar way for end users and integrates well with JupyterHub.
Q: I like the setup. Can I just use it as a personal workstation?
A: Yes! The tower hardware configuration would make a powerful desktop workstation for a single or multiple users. It is simple to add a full Linux desktop to the pre-configured server environment. The system is based on Ubuntu 18.04 LTS and the "tasksel" application allows for simple installation of several desktop environments. You still use Cockpit and JpuyterHub/Lab from the desktop.
Q: How secure are the servers?
A: The server configuration is based on Ubuntu 18.04 LTS and easily kept up-to-date with the Cockpit interface. The configuration is intended to be run on a local network (LAN) or accessed through a VPN. It is not configured for public network access. That would require a local network administrator to properly configure.
Both Cockpit and JupyterHub are served over secure https using self-signed SSL certificates. Traffic to the system is encrypted. The SSL certificates need to accepted into the users browser "trust store" on first use.
Q: Can I use something other than JupyterLab for my development work?
A: Yes. JupyterLab is a "standard" Machine Learning notebook based work environment. However, It is also possible to connect the MS VScode development environment to the servers and work with using the excellent VScode editor and tools. Also, the server can be accessed and used from SSH terminal connections if desired.
Q: What kinds of user authentication is available?
A: The servers are configured by default to use local user accounts and passwords. JupyterHub does support OAUTH, GitHub and Google authorization but these configurations do require setup by a skilled system administrator.
Q: Can I use other hardware configurations?
A: We are initially making the configuration setup available only on select systems.
Q: This is great but I really need more capacity will you be offering multi-node configurations?
A: The long term goal is to offer larger clustered configurations but no timeline has been established for this work.
Q: You haven't mentioned the framework that I want to use, can I get a custom configuration?
A:There will be a few standard frameworks configured by default. Instructions will be included to easily add other frameworks and languages. Setup details are provided to configure system wide resources as well as user personal account configurations.
Q: I have more questions who should I contact?
A: Please use the "Request a Consultation" form. We would love to hear your thoughts and answer your questions!
- Note: JupyterHub with JupyterLab Install using Conda
- Note: Self-Signed SSL Certificate for (local) JupyterHub
- Note: Auto-Install Ubuntu with Custom Preseed ISO
- Workstation Setup for Docker with the New NVIDIA Container Toolkit (nvidia-docker2 is deprecated)
Why Choose Puget Systems?
Rather than getting a generic workstation, our systems are designed around your unique workflow and are optimized for the work you do every day.
By keeping inventory of our most popular parts, and maintaining a short supply line to parts we need, we are able to offer an industry leading ship time of 7-10 business days on nearly all our system orders.
We make sure our representatives are as accessible as possible, by phone and email. At Puget Systems, you can actually talk to a real person!
Click here for even more reasons!