As of June of 2023, we are considering the JHPCE Web Portal to be in beta test mode. Please email any feedback or questions you have in using the Application Web Portal to bitsupport@lists.jh.edu.
We are excited to announce the launch of our new web application portal, designed to enhance your experience when accessing RStudio and Jupyter Notebooks on the JHPCE high-performance computing (HPC) cluster. This transition from a text-based SSH environment to a user-friendly web interface aims to streamline your workflow, provide greater accessibility, and offer a more seamless and interactive computing experience.
With the new web application portal, you can now access RStudio and Jupyter Notebooks directly through your web browser, eliminating the need for installing and configuring software on your local machine. This means you can dive into data analysis, statistical modeling, machine learning, and scientific computing projects more quickly and efficiently.
The web application portal offers a range of features to support your computational needs. RStudio provides a powerful integrated development environment (IDE) for R programming, enabling you to write, debug, and execute R code seamlessly. Jupyter Notebooks, on the other hand, allows you to create and share documents that contain live code, equations, visualizations, and narrative text, making it ideal for collaborative and reproducible research. We look forward to your valuable feedback as we continue to improve and optimize this environment to meet your evolving needs.
In order to access the Application Portal, you will need to be on the JHU network, either by being phyically on campus or connected in via the JHU VPN. Once on the JHU network, you can access the Application Sever from your web browser at https://jhpce-app02.jhsph.edu/
You will need to login to the App Portal with your JHED ID and password via the link in the upper right corner of the page. This will take you through the normal JHU/JHED login process.
Once logged in, you can click on the “Access JHPCE Apps” icon will be able to access the applications. You will see the screen below.
For both Jupyter and Rstudio, the process to launch the application will be identical. You can select the amount of RAM and number of cores you need from the dropdown lists in each application window. When you click the “Request” button, a new session of the application will be spun up on one of the JHPCE compute nodes. Once the session has started, you will receive an email containing a personalized link to use to access your session. Typically this will take 5-10 minutes to start up.
For Rstudio, you will be prompted for your JHPCE login and password, at which point you will see an Rstudio session on the cluster.
For Jupyter, you may see a screen similar to below, indicating an issue with the SSL certificate. We are using a self-signed certificate for JupyterLab, which is where the warning comes from. The actual warning page will vary from browser to browser, but you should be able to click on Advanced and then proceed to the Jupyter Notebook page. For some older versions of the Chrome browser, you may need to type the string “thisisunsafe” anywhere in the Google Chrome warning page to proceed.
The challenge in providing a web-based service like this in an HPC environment is that most software wants to run the application directly on the web server that is supporting it, and we want the application to run on one of our large compute nodes. In order to accomplish this we need to interact with the cluster scheduler to run on a compute node, and use a port forward on the external facing web server to bridge the connection.
For each of these service, a job on the JHPCE cluster will be submitted on your behalf. You’ll be able to see these jobs by running “qstat” on the JHPCE cluster.
[login31 /users/mmill116]$ squeue --me
JOBID PARTITION NAME USER ST TIME NODES NODELIST(REASON)
3256668 shared rstudio- mmill116 R 19:24:17 1 compute-094
As long as your job is running you will be able to access your instance of the Jupyter or Rstudio without having to relaunch a new session from the Web Application Portal. Please note that these jobs will be accrue charges just like normal JHPCE jobs (about a penny per hour for a single-core job using 5GB of RAM)
In addition to these Applications, you can also request a password reset, or a one-time Verification Code for logging in via ssh to the cluster. These features can be found under the JHPCE User Tools section from the Home page.
The architect behind this Application Portal is Adi Gherman, who has spent countless hours working through the intricacies of adapting these software packages to work in our environment.
*The 3 introductory paragraphs were largely written by ChatGPT with the request "We are moving some services on our HPC system from a text based ssh environment to a web application portal. The services include Rstudio and Jupyter Notebooks. Please write a 3-4 paragraph introduction to that environment that we can give to our end users."
