On Thursday, Mirantis announced a new initiative with partners Citrix, Metaswitch Networks and Overture Networks that will unlock telecommunications companies from expensive proprietary hardware, enabling them to launch new services with greater velocity and lower cost.

To find out more information and better understand what they are doing, I spoke with Kamesh Pemmaraju, Mirantis vice president of product marketing.

 

VMblog:  What exactly are you announcing that's new in NFV and OpenStack?

Kamesh Pemmaraju:  This week, Mirantis launched an NFV initiative - along with our partners, Citrix, MetaSwitch and Overture Networks. Our goal is to help network operators and communications service providers improve agility around their strategic business initiatives, reduce operational overheads, and innovate with fewer roadblocks. We're doing some critical new things. 

First, we provide a hardened Mirantis OpenStack platform optimized and fine tuned for hosting Virtual Network Functions (VNF's) that meets the scale, performance, and resiliency requirements of the communications industry.

Second, we're creating a partner ecosystem of VNF, orchestration, SDN, hardware and other pieces whose solutions are validated against this configuration of OpenStack. We're starting with joint announcements with three key partners who provide VNF and orchestration solutions:

• Citrix, for using NetScaler Control Center as VIM and NetScaler for virtualized load balancing and application delivery control, enabling chaining, scaling and dynamic fault tolerance for other VNF functions.
• MetaSwitch, who provide Perimeta vSBC - a virtualized Session Border Controller that performs SIP call setup/teardown and myriad other edge-network functions in advanced VoIP and multi-service carrier networks.
• Overture Networks, who have validated their Ensemble Open Service Architecture (OSA) - a complete carrier-facing service orchestrator for VNFs, SDN and other resources.

Overall, our goal is to make NFV on OpenStack practical, affordable and easy for carriers to consume, and to accelerate OpenStack becoming the standard Infrastructure-as-a-Service for the communications industry. 

VMblog:  Can you explain why NFV is suddenly so interesting to telcos?

Pemmaraju:  At its heart, NFV is all about virtualizing the network function hardware that previously has been expensive to acquire and operate as well as difficult to modify. NFV is really about cloud computing applied to the problem of how to provide agile, carrier-scale network functions, services and other communications applications. It's the end of a long process of replacing capital-intensive, monolithic, proprietary, relatively inflexible physical device solutions and replacing that with software running on VMs on open IaaS and commodity hardware.

VNF capacity scales out smoothly on demand and utilizes common infrastructure more efficiently, lowering capex. It mixes and matches nicely. Think about the difference between assembling a service by integrating and chaining physical appliances versus doing pipelining packets down a chain of software entities.  The latter approach is much more intuitively simple, smaller, and neater. And that's before you get to the operational benefits of NFV, where you can use automation tooling and web UIs to scale capacity, assemble new apps, and securely extend self-service capabilities to your customers.

VMblog:  To give us some context, can you describe a concrete customer example of an NFV deployment on OpenStack or describe the use case?

Pemmaraju:  Let's look at the use-case with Citrix, where we're deploying NetScaler Control Center to manage NetScaler ADC on top of Mirantis OpenStack. The service NetScaler is providing is adaptive load balancing - both on-demand and then adaptively scaling and self-healing - around VNFs that a carrier might want to use to provide some service: say, SIP call setup or IM or Presence.

That Load Balancing (LB) function is a primary enabler of the SIP app in several related dimensions. First, it lets you spin up an LB that serves traffic to a big enough set of VNFs to meet your SLA, and provides a kind of basic HA on this service - if one of those VNF instances is burdened, the LB guides traffic elsewhere to an instance with more momentary capacity. If a VNF dies, the LB routes around it until it can be put back in service, which can happen automatically with STONITH (one of my favorite acronyms: Shoot The Offending Node In The Head). If you lose an entire rack, or Availability Zone, no problem, because your app's VNFs are (of course) spread across several AZs (or data centers), so you have no single point of failure.

As your business grows and your customers get more sophisticated, you can add new VNFs, new LBs and new apps to utilize the same infrastructure to deliver multiple services. Your SLAs are all met because, in a worst-case scenario, if your app is architected right, NetScaler will keep your downside to what end-users will perceive as a small hitch in service as work is shifted to a new VNF from a failed one. This is way better than a generation ago with digital call-processing, when - if a rack went dark - you lost all the calls running on the rack (which was then considered a perfectly-acceptable loss, because it was contained).    

VMblog:  How is Mirantis' approach to NFV and OpenStack different from other vendors?

Pemmaraju:  At Mirantis, OpenStack is what we do best. That is our one and only focus. Because OpenStack is designed from the ground up to be open and flexible, we can adapt the platform to specific use cases which deliver massive business benefits to our customers and then we push those benefits upstream so the community at large can benefit from it.  This is exactly how we are approaching NFV -- we are strengthening, fine tuning, and configuring OpenStack itself for higher performance, resilience, and scale that carriers demand while also working on NFV-related but non-OpenStack projects such as OpenVSwitch and Intel DPDK.  All of this actually represents a quiet sort of revolution in itself, considering the way vendors have historically gone after the telco space, usually creating some degree of lock-in, and explaining that this lock-in was essential to assured service delivery.

The Mirantis NFV initiative will leverage the work we are doing at Ericsson, Telstra, Verizon, PacNet, AT&T and similar customers, building large scale clouds for production and understanding the unique requirements and challenges of the communications industry and in turn bringing this knowledge back into OpenStack. The partners we're working with themselves have vast domain experience in the communications service provider space and in many cases, that domain experience will ultimately improve the performance of OpenStack for the NFV use-cases and all its users.

##

Once again, a special thank you to Kamesh Pemmaraju, VP of product marketing at Mirantis, for taking time out to speak with VMblog.com.