Kumar Srikantan, CEO of Pluribus Networks, tells us what the company has been up to, how it’s different from the competition, and how Netvisor impacts the software-defined networking (SDN) and virtualization space.
SDxCentral: With your latest round of financing, you’ve raised $95M in total. What’s your plan for scaling in 2015 and beyond? What can we expect from you?
Srikantan: This latest round of funding is a key validation of our approach to advancing the SDN value proposition and provides us the fire-power required to run the play in a competitive and shifting landscape. This financing will help us to significantly raise our investments in sales, marketing, and engineering to take advantage of the opportunity ahead of us.
We intend to also build awareness and credibility for the company with this funding. And with customer traction increasing and deployments scaling, we are aggressively adding capacity in customer support and advanced services globally. Today we have about 75 full-time employees, and we expect to grow our headcount north of 50% over the coming year.
Can you explain some of the thinking around “SDN as a network application platform?”
A big part of the value of having a more intelligent, more programmable network is flexibility. Traditional networking involves dedicated hardware for specific functions like routing, switching, L4-7 services, and others, but the power of SDN lies in the flexibility of the infrastructure to enable new services and capabilities without constant hardware rip-and-replace based on ever-evolving infrastructure services needs.
Think of the good old world when a phone was a phone and a Walkman was a music player. It took multiple devices to get different services. The traditional network and services ecosystem of today operates in a similar way, with lots of dedicated, single purpose “vertically integrated HW-centric offerings” that must be integrated together for creating today’s data center infrastructure.
A key promise of SDN is to address the cost, complexity, operational manageability, and adaptability challenges of this traditional infrastructure with a forward looking “smartphone” approach that brings benefits of openness and programmability and solves the following key problems that matter to CIOs:
- Freedom of choice to source and deploy standard off-the-shelf HW platforms the way computing is sourced today,
- Extension of the IT operations model of computing to include networking, and
- Highly adaptable and flexible infrastructure (like computing world) that increases deployment and enablement of new network applications and services in a software-defined way.
OCP, ONS, off-the-shelf offerings in networking silicon, computing, and storage and the open source software ecosystem are enabling innovative startups like Pluribus Networks to build innovative platforms that are more like the “smartphone” in the networking world. The net result is a completely open and highly programmable platform that enables the network to deliver services and applications as dynamically as has been possible in the smartphone world.
How does the Pluribus Netvisor change the game in freeing CIOs from the current network equipment consumption model? How does a software approach change things?
In the past, functionality was tightly wedded to hardware. Need new capabilities? Buy a new box or a new line-card. Maybe you actually need a new line-card and a new supervisor engine. This is the legacy networking consumption model.
With Pluribus Netvisor, it’s all software and an industry standard-based open platform that was designed from the ground up to support virtualization and running third party apps as network services. This enables new applications and services to be rapidly added and deployed in an adaptable and agile way without repeated rip-and-replace of HW, as is the case with vertically integrated offerings from the traditional networking segment.
In our 2015 predictions we wrote about the emergence of repeatable SDN use cases. How do you align to these? What are the most popular use cases that Pluribus is deployed for today?
The Pluribus SDN platform supports a number of network services use cases that are applications on our platform. These include L2 switching, L3 routing, Fabric-Cluster Orchestration, Unified Virtual/Physical Analytics and Open Network Segmentation. Third-party applications include the TIBCO EMS and FTL messaging systems that are L7 applications riding on our platform. All of these speak to the power of the “SDN-as-a-Platform” architecture that enables a myriad of network services and applications to run on our SDN fabric via software.
A use-case of considerable interest is unified overlay/underlay network analytics. This provides comprehensive visibility, monitoring, and programmability control to applications in the network fabric for SLAs and security. At the recent Open Networking User Group summit in NYC, scale, visibility, and programmability in a unified overlay/underlay world were deemed top of mind requirements for the top 2000 Enterprises that are deploying private clouds. This use-case speaks to the heart of this need of Enterprise IT.
Network micro-segmentation is another key area of interest for many customers looking to enable private cloud infrastructures and services to their internal IT consumers. Our open protocol-based architecture micro-segments the network without custom HW. This enables customers to wire their infrastructure once, but then “re-wire virtually” when they want to bring in new tenants.
What’s unique about the Pluribus platform that allows you to make a difference in addressing these use cases? How did Pluribus’ deep expertise in server technology and distributed compute influence your approach?
We are bringing a server-centric view of the world to the network. First, with the Pluribus Netvisor, the network becomes a natural and seamless extension of computing. We treat the networking merchant silicon as a virtualized “HW offload” engine for all L2/L3 switching data-plane functions – exactly the way TCP Offload Engines and SRIOV technologies offloaded IO and data processing optimizations played out on servers.
By combining this powerful Pluribus Netvisor Operating System with open computing based server-switch-storage converged platforms, we create a far more powerful, flexible, and adaptable SDN platform for enabling new classes of software-defined-network services and network applications. This is done in a highly programmable way without having to constantly resort to infrastructure HW churn.
An example of a network-software-defined service is our fabric-cluster application that rides on top of Netvisor. Every Pluribus Netvisor switch on the network shares the same overall view of the network with every other Pluribus switch; one of the ways we are able to deliver unified physical/virtual analytics. The fundamental technology we use to accomplish this is based on traditional compute and database clustering and synchronization that have been deployed for decades in the computing world, now available as a “software defined service” in the networking world.
Another example of a software-defined-approach to solve a traditional networking challenge is one of scale in a highly virtualized data center. For instance, VM density increases are pushing the number of MAC addresses in L2 domains well beyond the 100-200k range that many current, traditional switches support. We get around this by taking advantage of the combination of networking, computing and storage in a way that we don’t depend on the physical limitations of traditional networking elements and can achieve far greater scale.
How would you categorize some of the industry approaches to virtualization and where do you see Pluribus fitting in?
It’s really coming down to two basic approaches – Cisco and VMware. Cisco’s network-centric approach with ACI is to bring programmability and agility into the network infrastructure. It scales, but is hardware-centric.
The VMware NSX approach, which is all server-endpoint or overlay centric approach, treats the network as a simple transport platform. It offloads all of the tunneling intelligence and processing for the overlay network onto the servers, with a central NSX controller providing orchestration. This works, but is not particularly efficient and you lose some visibility and control of the physical network. There are scalability and security challenges with this approach as the number of VMs scale in increasingly virtualized environments.
The biggest challenge with the above approaches is also that customers have to run a two-layer network – overlay and underlay — in an increasingly converged virtualized world. The Pluribus approach is premised on combining the best of network-centric and computing-centric approaches with network virtualization remaining in the network and scaling virtualization and compute to focus on business applications by keeping “computing clean.”
We keep network processing in the network fabric and take advantage of convergence of network-compute-storage to deliver innovative network services and network applications in a programmable way in the network as an SDN platform instead of general-purpose computing resources being taxed to run network protocols and services.
Computing resources are freed up to focus on business generating applications – as opposed to networking services. In short, our approach is a logical and scalable evolution of VMware NSX. It can also be viewed as an open alternative to Cisco ACI.
What about the white-box/brite box ecosystem? How is Pluribus different from some of the other vendors in the space?
Really it comes down to one thing: our OS, Pluribus Netvisor, is what makes us different. Unlike most other network operating systems, the Pluribus Netvisor is fundamentally built with best-of-breed open source software integration, kernel-based network virtualization that brings the full power of the network into the operating system, snd a high degree of server class programmability and orchestration from the computing world.
Our software-defined-network applications, like fabric-cluster technology, provides not only a single point of management but enables unique, deep insight into traffic on the network, allowing analytics to be done directly on the network. No need for taps, external network packet brokers, visibility fabrics or any overlay infrastructure.
We also bake bare-metal, type 1 virtualization capabilities into the OS, an approach that yields not only better performance, but also allows us to uniquely provide visibility into the underlay and overlay virtual network infrastructure all at the same time.
We were actually very flattered last year when we saw the Wedge announcements. The network guys at Facebook and OCP are incredibly sharp, so to have people like that validate our fundamental approach was a wonderful thing. In essence, Facebook is embarking on this evolution of infrastructure for their internal benefits. We are building on this vision and bringing the value to the broader enterprise market that is looking to build private cloud infrastructure for their internal consumption.