Migrating from Cumulus to SONiC for Broadcom ASICs

As a lot of people already know, Cumulus Networks was acquired by Nvidia/Mellanox in 2020, which was very exciting for some, but something of a problem for the customers who had Broadcom ASIC-based switches.

Now let’s be clear, Nvidia makes some of the most bad-ass switches on the planet right now. 10-400G in super compact form factors, perfectly optimized for all the new AI/ML workloads that need the sexy GPUs that Nvidia just happens to sell.  And if you want to get all of that Mellanox ASIC goodness (in particular their support for AI and storage networking is unparalleled), you’re going to stick with Nvidia.  Makes total sense to me, you cannot beat that sort of vertical integration.

So what if you are a Cumulus Linux user, but you absolutely need to use a Broadcom ASIC-based switch?  What are you to do?  In various other posts I have talked about why SONiC is fundamentally different from things like switchdev/DENT, so let’s put those options aside.

So, how can you prepare for a migration to SONiC?  What sort of learning curve and culture shock are you in for?

These are the major considerations when migrating from Cumulus Linux on Broadcom to SONiC:

Compatibility with existing hardware: One of the primary factors to consider when migrating to SONiC is the compatibility of your existing hardware with the new NOS. While SONiC supports a wide range of hardware platforms, it is important to ensure that your current hardware is supported to avoid any potential compatibility issues.

Features and functionality: Another critical factor to consider is the features and functionality offered by SONiC compared to Cumulus Linux. While both platforms offer similar capabilities, there may be slight differences in the way features are implemented. It is important to evaluate these differences to determine if SONiC will meet your organization’s requirements.

Migration strategy: Migrating from Cumulus Linux to SONiC requires careful planning and execution. It is important to develop a migration strategy that takes into account factors such as downtime, data migration, and potential disruptions to existing network services. It is also important to test the new network infrastructure thoroughly before deploying it in a production environment.

Training and support: The migration process can be complex, and it is important to ensure that your team has the necessary training and support to manage the new network infrastructure effectively. This may involve providing training on new tools and features, as well as ensuring that your support staff has the necessary expertise to troubleshoot issues that may arise.

Cost: Finally, it is important to consider the cost implications of migrating from Cumulus Linux to SONiC. While both platforms are open-source and free to use, there may be additional costs associated with hardware upgrades, training, and support. It is important to evaluate these costs carefully to ensure that the migration is financially viable.

I would argue that the above list is a smaller challenge and learning curve than you initially experienced when moving to Cumulus.  So if you’re already familiar with open networking solutions and how to build and operate disaggregated network equipment (i.e. open networking), you shouldn’t have much trouble.

P.S. An upcoming post will dive into the technical details of the migration tasks.