Netvisor Unifies the Physical and Virtual Network

Just over a year ago we announced Virtualization Centric Fabric (VCF™) as the architecture underneath our family of Netvisor OS offerings. Pluribus was founded on this architectural approach to networking with the fundamental goal to take advantage of the power of open and commoditized switching hardware. As it turns out, the switching silicon that has been created in the last few years can actually run circles around the vast majority of the proprietary chip sets being developed by the traditional vendors, and when coupled with a software-defined feature set, end-users receive all the benefits.

VCF is the corner stone of our approach to unifying the physical and virtual network. VCF provides a clear business advantage and  major architectural shift away from legacy proprietary network operating systems and their rigid networking topologies. As modern business service application PODs (typically a collection of hundreds of servers installed in 4-20 racks and running virtual machines and/or containers, all connected via a leaf/spine switch) become more prevalent (and actually becomes the unit of measure of an application building blocks), the need for running the virtual and physical network as one manageable entity has become critical. The need for multiple managed approaches for various portions of an infrastructure is regularly cited as the number one customer pain point and the number one inhibitor to business agility. This pain point becomes more amplified as users begin to seek agility: by forming their private clouds, Hybrid clouds, or fully articulated Software Defined Data Centers (SDDC).

Pluribus Network’s VCF architecture solves this agility problem by fundamentally changing the architecture of the operating system installed on each switch to align more closely with the proven approach seen with server virtualization. VCF virtualizes switching resources, and then allows those resources to be applied to specific business needs. It shifts resources from consumption as physical ports to virtual ports (“vPort”); it replaces H/W tables with distributed S/W tables, and adds a high performance distributed control plane which interacts fabric-wide with all of the involved switches. Pluribus VCF virtualized the network, eliminating the hardware boundaries we have all grown up to hate.

In specific, Pluribus VCF allows:

  • Physical switch ports to become a “vPort” – A fundamental tenant of legacy networking is each physical server is tied to a physical switch port and a hardware switching/routing table maps that connection. The Pluribus Netvisor operating system redefines that end-point using the concept of a virtual switch port that represents any server VM, Container, discrete or Mobile device. In this way, any unique combination of MAC address and VLAN/VXLAN is treated as an index into the vPort distributed hash table database and the switch chip hardware tables located in every switch appliance are updated based on this singular global view.
  • H/W tables to be transformed into Distributed S/W tables – Today’s volume switching chipsets are very powerful and can automatically learn and update a simple switching table. But by doing so, the switching function would be limited to just the local information and not a global view. This is in sharp contrast to the world of servers where the kernel controls every aspect of server CPU and its associated network interfaces and maintains all of the tables of network addressing. Accordingly, each Pluribus Netvisor operating system running on any open switch keeps its switching and routing table in S/W that it maintains, and the switching chip tables are used as essentially a cache. Any time the hardware has a cache miss, that is raised to the Netvisor operating system as a trap which in turn looks up the miss in its distributed virtual port (vPort) table to figure out if it’s a VM migration, new VM/Container coming up, a mobile device getting active or this a misconfiguration/loop or security breach.
  • Traditional single switch control planes to be transformed into a high performance and robust distributed control plane that is clustered with all other switches over the high-speed network itself. Most modern open switches have amazing amounts of processing and I/O capabilities. These high capacity processor configurations allow the Netvisor operating system to move all learning and misses to software via the high speed PCI-Express and then update hardware tables that are directly memory mapped over PCI-Express in software. The vPort database is distributed across all Netvisor instances in a leaf/spine pod. This hardware-leveraging architectural approach can easily span and manage millions of vPorts, VM migration, VM/Container life cycle analysis, analytics and security policies.

In essence, Pluribus Netvisor transforms  the switch chip hardware-based end-point table to a software-based distributed vPort table which identifies the vPort current location based on a physical switch/port in the fabric. And like any higher-order distributed and clustered processing system, Pluribus Netvisor also injects the programmability and API access into the switching layer. And since the Netvisor VCF technology is fully distributed, changes made from any Netvisor instance are propagated across the entire switching fabric.


About the Author

Sunay Tripathi

Sunay Tripathi

Sunay is the CTO and a Co-Founder of Pluribus Networks. Prior to Pluribus, Sunay was a Senior Distinguished Engineer for Sun Microsystems, and was the Chief Architect for Kernel/Network Virtualization in Core Solaris OS. Sunay has an extensive 20+ year software background, and was one of the top code contributors to Solaris. Sunay holds over 50 patents encompassing network and server virtualization.