Alcatel-Lucent (ALU) launched its spin-out Nuage Networks last Tuesday (April 2) with great fanfare, and to positive response from press, analysts and network operators. In creating an end-to-end vision for network virtualization encompassing the cloud datacenter and the WAN, Nuage blazed the trail for best practices on how an incumbent networking vendor should enter the SDN marketplace.
SDNCentral had the opportunity to sit down with Sunil Khandekar, CEO of Nuage Networks, and have Sunil answer a list of questions from our members who were chomping at the bit to understand more about the company’s launch and what it means for the SDN space.
SDNCentral:Thanks for taking the time to speak with us, Sunil! First of all, everyone wants to know, why a spin-out? Why not do this as just another product line within Alcatel-Lucent?
Sunil: “To us, a spin-out made sense for this new market. We wanted complete independence from the mothership and looked to start this with a clean-slate. Within Alcatel-Lucent, we would be forced to look at the problem from the view of the installed-base, rather than taking a fresh look from the IT viewpoint and asking ourselves what needs to be changed.”
SDNCentral: So, how did you go about making the spin-out happen?
Sunil: “We found and recruited a really good team that already had strong networking DNA, and also drew from the IT side—members that knew the IT space well. We brought them together under a separate structure which gave us the autonomy, agility and the ability to attract the right talent. For the record, spin-outs are not entirely new to Alcatel-Lucent. Under Basil Alwan’s (Basil is the president of ALU’s IP division, and was the founder of TiMetra Networks, acquired by ALU in 2003) stewardship, we have done this in the past—in the security solutions area and another recently in the white-label cloud arena for gaming (CiiNOW). For this particular spin-out though, we structured it as a fully-owned subsidiary of ALU, allowing the team to focus on the product, while leveraging the intellectual property and assets of ALU, as well as the go-to-market marketing team and sales channels. We will also certainly benefit from ALU’s support and services capabilities.”
SDNCentral: Thanks for the insight into how Nuage was formed. Perhaps we can spend some time on the Nuage VSP (Virtualized Service Platform) solution and what problems you set out to solve.
Sunil: “We set out to solve the problem with network virtualization within the datacenter. We wanted our solution to focus on the three basic tenets of: (1) abstraction, (2) automation and (3) unrestricted and boundaryless networking.
The goal is to create a fully virtualized multi-tenant network that meets the needs of enterprises and service providers today. It’s not about SDN or OpenFlow or what protocol to use, but about solving the business problem today in a way that reduces complexity for our customers and reduces their overall cost and increases their agility. Our VSP automates the creation of multi-tenant networks within the datacenter and provides an abstraction that handles layer 2, 3 and 4 services. Furthermore, our VSP offering provides for integration with MPLS in the WAN, facilitating communication of our virtual network with private clouds or with remote locations. This is the most comprehensive SDN solution that we are aware of being offered today.”
SDNCentral: What are the components that make up this VSP solution?
Sunil: “The VSP platform consists of a policy management and analytics element (VSD or Virtualized Services Directory), a Virtualized Services Controller (VSC) and the VRS (Virtual Routing and Switching) agent, an enhanced version of Open vSwitch that acts as the intelligent forwarding and encapsulation/de-encapsulation element on the hypervisor.”
SDNCentral: What infrastructure needs to be in place in terms of hypervisors, compatible ToR (top-of-rack) switches or routers? Does this only work with ALU equipment?
Sunil: “VSP will virtualize and automate any existing data center network independent of the network infrastructure. VSP is agnostic to the datacenter network equipment, as it is to servers and the compute virtualization environment. We also work with all the major hypervisors such as KVM, XEN, VMware and Hyper-V, and have different versions of the VRS available for the different platforms. Since we use VXLAN as the encapsulation across hypervisors, we are not dependent on a specific type or brand of ToR switches to function. For integration with edge routers, we use MPLS over GRE, which provides maximum compatibility with existing infrastructure.”
SDNCentral: What are the best fit customers for the VSP solution?
Sunil: “Since the VSP platform is flexible and has limited dependencies on physical infrastructure, it works well in both cloud service providers and in enterprise datacenters. Any customer looking for multi-tenant network with automation and programmability is a good fit. With our ability to integrate the WAN into the picture, we can uniquely solve private to public connectivity for hybrid cloud deployment or inter-connections between private clouds.”
SDNCentral: In using VXLAN as the encapsulation, do you run into scalability challenges or dependencies on IP multicast that we’ve been hearing about?
Sunil: “Actually, we don’t. We are not dependent on having IP multicast support in the switches to enable L2 communication between VMs. We use MP-BGP to distribute reachability information for L2 and L3. The VRS elements are programmed with appropriate L2 and L3 forwarding entries along with L4 classification rules. This eliminates the need for broadcast and unknown traffic.”
SDNCentral: Does the VSC orchestrate the setting up of virtual networks in the VRS based on templates and policy in the VSD?
Sunil: “We prefer using the word ‘choreograph’ to describe our solution. We use our VRS agents to notify us when the environment changes—such as when a VM is instantiated, destroyed or moved. Our event-based model allows us to efficiently enable VM mobility based on detecting a state change, regardless of which orchestration platform initiated the change: CloudStack, OpenStack, vCloud Director, or SCVMM. Once we detect a change, the policies in the VSD direct the VSC to take the appropriate action based on the service template and policies to rapidly connect the VM into the right service. The same mechanism is leveraged to enable true VM mobility.
Our event-based model allows us to quickly integrate into a wide-variety of cloud orchestration platforms and hypervisors, unlike other solutions that have to manage the complete orchestration of virtual networks.”
SDNCentral: What protocols do you use to perform your magic? OpenFlow?
Sunil: “For us, it’s not so much about protocols, but solving the right problem with the right protocol. We do use OpenFlow, but also use XML, SNMP, MP-BGP, XMPP and others to get the job done. If we do our job right, the protocols become secondary.”
SDNCentral: What customers do you have running this today?
Sunil: “As it turns out, the platform that the VSC is based on is running across hundreds of service provider environments today—it is a very robust routing engine. The VSP solution is being rolled out for trials across a number of customers, including UK cloud service provider Exponential-e, French telecoms service provider SFR, Canadian telecoms service provider TELUS and US healthcare provider, University of Pittsburgh Medical Center (UPMC). We will be starting trials in May of this year and expect more roll-outs through the rest of 2013.”
SDNCentral: Thank you for your time, and we’re looking forward to hearing about how the early deployments are going in a few months!