Frankly, I came into the Pluribus presentation with extremely high expectations and the presentation did let me down a little. The technology is great and has significant potential but I am now questioning the direction and execution. I feel that Pluribus is suffering from a desire to ‘boil the ocean’ approach. Fundamentally they have lofty goals of unifying the Data Center under on common abstraction layer. If they are successful in doing so this would indeed be a huge achievement but I am not sure they can execute on all of their goals. It is not a problem of not having the core technical competency or vision but simply a problem of resource limitation versus execution complexity. For clarity here is the mission statement put forth by Pluribus in a solution white paper (http://www2.pluribusnetworks.com//l/36732/2014-06-23/lh5p):
“The goal of Netvisor is to deliver an open, programmable Network Hypervisor to bring both the virtualization revolution and the disruptive economics of the server industry into the networking world.
Netvisor is the first data center operating system, which solves the complex problem of virtualizing the infrastructure at the physical network fabric layer, which is the centerpiece of the data center infrastructure. By treating the merchant silicon switch chip as a true extension of the server, Netvisor is the foundation of a highly converged and virtualized architecture fusing compute, storage and network under one operating system for the Data Center.”
The NetVisor (which stands for Network Hypervisor) OS has always been the key to their solution and it is very versatile solution. In short it provides a Cumulus like Linux NOS but one of the key differentiators is what Pluribus refers to their Netvisor OS as a “Distributed Network Operating System”. Each Netvisor OS instance is synced with other Netvisors in the same fabric such that you can configure any device in the fabric from any switch.
The primary overlay transport technology is predictably VXLAN. This proprietary dynamic network fabric combined with a white box capable OS is the key building block for large number of potential use cases but this is where the story seems to devolve. Instead of focusing on key use cases Pluribus seems to be going after all possible use cases and their leadership seems to be unwilling to admit that their may be any shortcomings with any portion of the solution.
This was made evident in both Network Field Day 8 and 9 in which questions were presented that were never answer. In NFD 8 the question of a distributed gateway was mostly ignored and glossed over but never actually answered because the capability didn’t exist. In NFD 9, the simple suggestion that there may be a flaw and some FUD slinging in statements about server hypervisor overlay scale sent the CTO into a tirade. The result is that for a second straight NFD the delegates left Pluribus with unresolved discussion. For example see the following blog details from Ivan Pepelnjak (@ioshints):
In my mind it is not a matter of if Ivan or Sunay Tripathi (@SunayTripathi) is correct. Frankly, I do not know and it would be arrogant of me to state that I know enough about this topic to say definitively one way or other. I will say that in theory I tend to side with Ivan on this matter especially when you take into account development efforts around initiatives like Intel DPDK (http://www.intel.com/content/www/us/en/intelligent-systems/intel-technology/packet-processing-is-enhanced-with-software-from-intel-dpdk.html). If Intel has anything to say about it, I have to assume they will figure out the problem when all is said and done. The bigger issue is the fact that when challenged they seem unwilling to listen and to have a frank discussion. It could have just been a heat of the moment situation that I completely just misreading but this is the second NFD in a row that something similar has happened.
On the other hand, I do not mean to say that Pluribus does not have a good technical story, on the contrary it is very significant but the humility that no one person or company can possibly achieve everything seems to be lacking. That is my single largest concern now. What additionally exasperated this impression was the candidacy of the vendor we had just departed from. Cumulus knew who they were, they knew exactly what they were passionate that they could achieve but they were equally as candid with what they could not.
In summary, Pluribus has some excellent technical capabilities, I just hope that I simply have made an inaccurate read on their direction and leadership. On a plus side I am definitely still very excited about their technology development and will certainly continue to track their progress.