Celebrating Pi Day with Graphistry and Kasm Workspaces: A Breakthrough in Collaborative Data Analytics

Just like any good story, this one begins with a group of friends discussing their challenges regarding work, life, and everything in between. During our conversation, we stumbled upon a topic that many enterprises, universities, and small businesses face: rising costs of securing data, limited access to HPC computing resources, and increasing vectors of data sharing. It just so happened that I was hanging out with a couple of data scientists in the Bay Area that night, and we began discussing Graphistry. If you're not familiar with Graphistry, it's the most scalable graph-based visual analysis and investigation automation platform on the planet. I had come across it last year in my AI class while working with Jupyter notebook for my final, and Alexander Morrise, PhD who went on to become Graphistry's Head of Graph Data Science helped me become familiar with it.

#Graphistry provides visual graph intelligence to your big or complex data, automatically transforming your data into interactive, visual maps built for the needs of analysts. It is commonly used in cybersecurity, Fintech, Federal, Social networks, and Dev and Data Science - the same areas that I serve at Kasm Technologies with our Enterprise CDI platform,?which delivers streaming containerized apps and desktops to end-users. The Workspaces platform provides enterprise-class orchestration, data loss prevention, and web streaming technology to enable the delivery of containerized workloads to your browser. I felt like the old "peanut butter and chocolate" senses were going off, and we could address those three problems together in a way that would benefit end-users and hopefully support a path to solving the challenges presented above.

After talking with Graphistry's founder Leo Meyerovich a few days later, we found that CDI could help with Graphistry,s Users' deployments in three ways:

1. supporting a weak client,
2. supporting a weak network,
3. and improving collaboration and DLP (Data Loss Prevention).

While most Graphistry customers are happy with the standard experience, occasional extreme environments do come up, such as in government agencies with heavily locked down client devices, utilities with physical network separations, and students who need to make each dollar goes the extra far. This solution could fit a lot of use cases, not just for Graphistry users, but many deployments where data analysis was needed in industries like finance, healthcare, and more. These three conditions need to be met at a minimum for compliance and productivity to exist.

The Experiment: So, we began our experiment by removing some of these barriers to insight and innovation on a mass scale. We believe in the community as an open-source project, just like Graphistry, which offers a community edition at the Graphistry hub. As well as popular open-source libraries like PyGraphistry. I remember a famous story of a boy who got access to a PGP (HPC of its time) with his mentor. They ended up starting a revolution I'm writing in a word processor that was that boy's legacy. How could we not try to tackle this challenge and make a difference?

To start the experiment, we utilized the Kasm Workspaces by downloading it from here. After deployment, whether on-premise or in the cloud, we built a Graphistry container and a custom image?using the Chrome Core container?that pointed us to hub.graphistry.com for the container app. Additionally, we integrated the Jupyter desktop into the Kasm Core Ubuntu custom image for the Graphistry Dev Desktop. On the Kasm Workspaces Agent, we configured the NVIDIA GPU?and enabled it for both the Graphistry App and Graphistry Desktop. Once that was completed, we wanted to test the Chromebook as a weak client. We deployed the Desktop and App using Kasm Workspaces' PWA?(Progressive Web Application) to the Chromebook and started comparing a GPU-enhanced session on a Kasm Container App to Graphistry on a Chromebook native. The side-by-side comparison showed some initial relief, but running hotter and navigating large data maps proved challenging. The Kasm container had a significant advantage as it did not experience the "weak network" effect, and the endpoint was not competing with hundreds of devices as we were experimenting with shared WIFI networking at the AWS Startup loft in SF, which was crowded that rainy day. We made some real progress in this initial experiment, which led us to wonder: how low could we go on the endpoint? How close can we get the data to the browser? How far could the endpoint be if we got the rendering browser closest to the data and leveraged Kasm's streaming and rendering technology to the endpoint?

?Testing client limits with Raspberry Pi

The Fringe scientist in me wondered if we could use a Raspberry Pi easily in this situation. Would it be worth the exercise, given that many people tried using the Raspberry Pi as a work-from-home device during the pandemic? So, I dug out a couple of Raspberry Pis (come on, you know you have some too!) and wanted to see how far they had reached in the enterprise since their introduction, leveraging VDI. Could this technology also be leveraged in CDI? Quick shot out to those not familiar with CDI versus VDI. Container Virtual Desktop Infrastructure is an enterprise-class orchestration, data loss prevention, and web streaming technology that enables the delivery of containerized workloads to your browser. Whereas VDI or DAAS work by leveraging presentation protocols to stream virtual machines or virtual apps to other desktops. CDI is delivered through Docker containers, dramatically reducing platform resource requirements and enabling sessions to boot in seconds, rather than minutes. In this context, we wanted to deploy to a Raspberry Pi as the endpoint for one of the most advanced workloads. The Raspberry Pi has promise as a VDI device, but it faced challenges, including a lack of dual monitor support and management issues. The enterprise-grade Raspberry Pi overcame these challenges, offering virtual access securely and at a lower price point than traditional thin clients. The Raspberry Pi 4 further established it as a viable alternative with next-generation capabilities. Stratodesk's workspace?hub was a breakthrough towards an enterprise-ready Pi device, offering day-one support for the Pi 4. So, I called up my good friends at #Stratodesk and started working off my Pi with a Chromium Browser connection to my Kasm Graphistry App, and we saw some amazing things. A less than 50 USD device could access a cybersecurity data visualization over a low-bandwidth/low-end endpoint.

The Results: Now we have achieved a weak endpoint that can be leveraged via CDI to support the three major challenges:

Weak Client: We leveraged both a Chromebook and a Raspberry Pi to access Graphistry Hub with high fidelity. GPU-enabled Chrome sessions support higher render rates with less session support handling by leveraging Kasm Workspaces container streaming and rendering tech.

Weak Network: We leveraged both a local connection that was crowded like a university WIFI or shared workspaces for the endpoint, or more importantly, we have it running in AWS, and you can deploy Graphistry from the AWS Marketplace to leverage it on the same VPC as the Kasm Container Browser and let Kasm support the session handling.

Collaboration and DLP: The Graphistry App is a Kasm Core Chrome container that is in the least Privileged Container that is deleted after use. This container also has session controls like uploads, downloads, and even sharing to enable collaboration, as demonstrated in the recordings.?

So Happy PiDay! This experiment went extremely well, and with little more than documentation to set up, this could be used in a STEM classroom, a NOC facility for Cyber investigations, or even pharmaceutical labs to manage protein folding. The ability for a startup to consumer curiosity barrier to be lowered with access to working with some of the most advanced workloads. Open-source and community projects are open for both Graphistry and Kasm Workspaces.?As you look towards startups, enterprises, and universities, there are also SAAS implementations of both Kasm Workspaces and Graphistry to enable your data analysts and data scientists that could be leveraged on a subscription basis.