As a Network Engineer the ability to adapt to changing environments and evolving technologies is a critical aspect of our profession especially when faced with customers always shooting for lower expenses, improved capability and performances. That has been my case few months ago when I was introduced to NVIDIA Cumulus Linux. In this article series, we will try to understand why 34% of fortune 50 are embracing a product which has been around since 2010.
What is NVIDIA Cumulus Linux?
It is just another Linux Distribution
First of all, let me make it cristal clear: NVIDIA Cumulus Linux is Linux. In fact, NVIDIA Cumulus Linux is a distribution based on Debian. In other words, if you are familiar with Debian OS or its most popular derivative, Ubuntu, you should be confortable navigating the NVIDIA Cumulus Linux OS.
Agile & Extensible
NVIDIA Cumulus Linux is highly customizable. 90% of NVIDIA Cumulus Linux code is in python so you can investigate and modify what you want directly from the switch as a consequence, we have quicker bug fixes and continuous code optimization. Also, since access to the code is open to everyone, it can be edited and rewritten in a faster language such as Go (This would make a very interesting project).
platform agnostic
NVIDIA Cumulus Linux is the best open network operating system currently on the market. Compared to Cisco IOS or NX-OS which can only be run on Cisco hardware or Junos on Juniper hardwares, NVIDIA Cumulus Linux offers a vast choice of hardware to choose from. In fact there is a list of recommended hardware with proven compatibility.
Architecture
Linux distro for networking
NVIDIA Cumulus Linux distinguishes itself from other Linux distributions with its focus on networking. switchd is an example of proprietary service at the heart of Cumulus Linux. It communicates between the Hardware and NVIDIA Cumulus Linux OS, and all the applications running on NVIDIA Cumulus Linux. There are other services specific to Cumulus that we will take the times to dive into in our future posts.
Why Would a company adopt NVIDIA Cumulus Linux?
The Future of the Datacenter
Merchant Silicon is already dominating the datacenter ecosystem. More than 63% of all switches in the datacenter around the world are using some form of merchant Silicon and the number is growing. Therefore companies should decide whether or not they want to remain tied to a single vendor by using an OS which can only operate on a specific hardware or else they want a OS that can be deployed on almost any hardware.
Cost
Most people don’t know it but when you buy a laptop with Windows OS installed, you are paying for both the hardware(Dell, Lenovo, HP, Acer, etc…) + OS (Windows 11). Therefore you can decide to buy just the hardware without and then install on it a free OS like Ubuntu Desktop. This analogy is a simplify way of explaining CapEx saving. In other words, Reduced CAPEX can be achieved by lowering the cost of acquiring open switches and optics.
Similarly, OpEx saving is derived from the freedom of choice across 100+ recommended hardware platforms. You can basically choose a hardware based on your needs and budget. It is estimated that customers have achieved operational efficiency and reduced their time to production by as much as 95%, while reducing their TCO by up to 60%. With rich ecosystem and converged administration, Cumulus Linux helps to get simpler, scalable, faster-to-deploy networks.
Simplicity
With Cumulus Linux, customers can run their data center networks the way Facebook, Amazon, Apple and Google have done for years, highly automated, without all the development time or expensive, specialized hardware. With automation as a layer of abstraction, System Engineers can now execute routine Network Engineering task with great efficiency while Network Engineers focus on designing and optimizing the network.
Summary
Although it might seem that I advocating for NVIDIA Cumulus Linux against market leaders like Cisco and Juniper, I would beg the difference that I am merely acknowledging the future of Networking and with the evolution of my tribe from Network Engineering to Solutions Engineering through DevOps Engineering.