Pure Play NFV: Lessons Learned from Masergy’s Virtual Deployment for a Global Enterprise

Network functions virtualization is just getting off the ground, so intelligence on deployments is vital news as the telecom industry moves towards these advanced virtual and software-defined networks.

One cloud provider to enterprises who is making a stir in virtual technology waters is Masergy, a private firm based in Plano, Texas, that competes with the likes of AT&T, Verizon, and CenturyLink.

Significantly, in mid-2015 Masergy announced the global deployment of an NFV solution for a not-yet-named Fortune 500 firm.  The Masergy Virtual f(n) acts as an app store from which customers dynamically choose services.

The software vendor supplying pure-play virtualization platform that enables Masergy’s solution is ADVA Optical Networking, (formerly Overture Networks, which was acquired in January) and the Chief Technology Officer of its new Ensemble Virtualization division, Prayson Pate joined us to provide an excellent review of developments in this important case.

	Dan Baker, Black Swan Editor: To start off, Prayson, can you give us a quick backgrounder on Overture?

Prayson Pate: Sure, Dan.  Overture got rolling in 2000.  Our focus early on was providing Carrier Ethernet solutions — both at the customer prem and aggregation point.

Currently we are selling into Tier 1, 2, and 3 accounts around the globe.  We supply to five of the top seven global Ethernet providers, and also eight of the top ten in the U.S.

Starting in 2012, we began to reinvent ourselves.  We transformed from hardware devices to supplying more complete solutions.  Now a big part of that is evaluating and deploying NFV and SDN.

We were just acquired by ADVA Optical Networking, helping make ADVA into a NFV powerhouse with a very broad portfolio of WDM, Carrier Ethernet and synchronization products.  Together we have a very broad base of customers around the world.

Our overarching goal is to help our customers improve their service delivery.  It’s about helping them put the speed into service deployment, taking the costs out, and driving more revenue.

	And what would you say the main benefits of NFV are for operators?  What do you get with NFV that’s better than what they have today?

NFV gives telecoms the ability to fully take advantage of virtualization, cloud, and data center technologies.  And the trend toward what we are calling the Telco Cloud is being driven by three things:

• Server flexibility — The operator can choose any kind of server it wants across a variety of price points and performance.
• Hardware isolation from software — It’s great when you can assemble software components without worrying about the hardware.  It promotes agile development and DevOps deployments.
• Managed connectivity — What’s missing from cloud and virtualization is connectivity.  And, of course, connectivity is exactly what telecoms have delivered to enterprises for decades: things like reliable access, QoS, and assured bandwidth.  So NFV allows you to map network functions like routers and firewalls with full attention paid to these performance and throughput issues.

	Excellent, so please give us a deep dive on Masergy’s virtualization case with an enterprise client.  What was this case about?

In June of 2015, Masergy deployed a virtualization on the customer prem to enable their Fortune 500 customer (based in the US with sites around the world) to dynamically request and order services from a customer portal.

So no longer does the customer have to call up and order services that arrive in 30, 60, or 90 days.  They can go on a portal, see how much a service is going to cost and order it dynamically.  This case shows how you can bring the whole cloud and instant delivery model to the telco world.

The initial set of services Masergy deployed for the enterprise was router and firewall offerings.  Several more services are in the pipeline such as a light firewall, encryption, and WAN acceleration.

	Why did Masergy decide to go the virtualization route?

Certainly delivery issues were slowing their business so they were desperate for solutions that delivered fast time-to-market.

Another is the problem of getting equipment to a wide variety of locations around the world.  Masergy’s CTO Tim Naramore said “it’s okay if the global enterprise needs to send a truck one time in a place like Mongolia or Patagonia.  But it’s not okay to send out a truck every time you add or change a service.”

This is why they wanted a virtual solution delivered through a single software platform across Carrier Ethernet.

And furthermore, Masergy asked to get it a pure software platform giving them the ability to make changes at will and enable offices in different regions of the world to adapt the solution to the local needs.

So Overture fit the bill with our pure software implementation of NFV/SDN.

Let me briefly walk you through some of the details of how Masergy deployed their solution.  There are some important lessons learned here that other operators will appreciate.  So here are five key philosophies that have guided their early success with NFV:

Masergy’s Strategy for Successful NFV Deployment

1.  Adopt NFV using a “Pure-Play Software” Design

Masergy considered going with a hybrid approach to virtualization that combines a separate Ethernet access device and a compute element.  But they opted against that course because it would have tied them to particular hardware and provide less flexibility.

They selected ADVA’s pure play software design.  And what we did was virtualized our Carrier Ethernet so our customers could use their preferred Intel ATOM-based server.  In this way there is no Ethernet hardware — no NID component.

What Masergy gained was complete flexibility on the hardware they choose.  For instance, if they need a bigger server, they can go buy a bigger server.  If they need a hardened server, they can get that.  In this case Masergy is deploying on the ADVA 65vSE, which is an Intel-based server wrapped in a carrier-class chassis.

So we gave them the ability to completely decouple the hardware from the software to enable maximum choice.

2.  Select Servers to Optimize Sourcing, Reliability, Costs

A key constraint Masergy had was the need to efficiently deploy servers in countries all over the world.  And a key challenge there is import/export certifications, which take an inordinate amount of time.

So instead of trying to get a piece of equipment into a country, with a pure play NFV solution, Masergy could now buy a local server and run on that.

So Masergy has lots of deployment choices today.  It can offer highly available servers using two routers connected to access links.  Or alternatively their customers can replace that with two servers running two instances of a router.

Topologically they look the same, but now, instead of stacks and stacks of boxes, you can have a pair of boxes that fail over to one another at the hardware level.

And if the enterprise opts for a more economical solution, they can just get a single box.  In either case — the cheaper route or a fully resilient high availability — you can get it running using standard servers.

3.  Create an Ecosystem of Suppliers

Masergy’s virtual platform is supported by multiple suppliers.

ADVA supplies the accelerated vSwitch (Ensemble Connector) that is fundamental to achieving the performance at the network edge.

Our Harmony partnership program includes three kinds of suppliers.

First are the VNF partners.  Masergy is using Brocade for virtual routers and Fortinet for virtualize firewalls.  Other VNF suppliers are being added soon.

The second group of suppliers in the Harmony ecosystem is the infrastructure suppliers.  Those include hardware suppliers like Dell, SuperMicro, and software players like RedHat and Wind River.  The final group is integrators including Kapsch in Europe.  Some carriers are looking for integrators to put these pieces together and deliver it.  Of course, very large carriers like Centurylink and Verizon will do that all by themselves.

4.  Work with Customers and Suppliers as Partners

Masergy is getting benefits beyond the speed and cost savings of deployment.  Another key advantage with the pure play software approach is they can achieve agile development and DevOps capabilities.

Virtualization also fundamentally changed their thinking around working with their customers and suppliers as true partners.

For example, Masergy compensates its team based on customer satisfaction, which has led to high customer ratings.  That’s achieved not just through staying in touch and keeping the services up and running.  In fact, Masergy developed a much higher appetite for doing custom work than your typical operator.

Masergy has made major investments in their systems to enable customers to not only get a portal to watch the status of services and billing, but also to actually make changes to their service.

On the supplier side, Masergy makes sure that their business model provides adequate rewards and incentives, which keeps customers loyal.

5.  Drive a New Commercial Model with Suppliers

Traditionally a telco supporting enterprises spends millions of dollars to buy a bunch of hardware and software that takes years to integrate and install before you generate revenue.  Those days are over.

Masergy, like so many service providers, is eager to push the cloud service model: order what you want when you need it and pay as you go.

So shared risk and share reward models are valid strategies.  And so is “we don’t start paying you for your functionality until we start billing for it.”

The Ensemble architecture includes a component called Registrar, which specifically facilitates these new flexible licensing models.

	Prayson, this is a highly interesting case.  Now, I’m curious, when you first approached Masergy and other customers about your “pure play virtualization” capability, what reaction did you get?

Well, as you can imagine, customers are quite skeptical.  And this is because they are used to lots of handholding from their suppliers.  But we sit down with them and walk through it all: openness, ecosystems of suppliers, pure software, and running on standard hardware.  They are naturally cynical.

It’s takes time for reality to sink in.  But after a while — and once they do it themselves — it becomes very appealing because this is exactly what they’ve been seeking for a long time: great flexibility and control.

And the test is simple: can you run your VNF on a standard server?  When we do performance tests, we run them not on our own server but on a SuperMicro server to prove this is not smoke and mirrors.

We host these functions in a standard server on your customer prem without any hardware assist and that’s powerful.  On the orchestration side, they see you can orchestrate a wide variety of VNFs from a wide variety of suppliers and some open source distributions as well.

When they ask if we really build a system that is composed of components from a variety of suppliers, and including some open source components, we emphatically say, “Yes, and here’s how it works and we’re happy to show it to you.”

Now it’s not ADVA alone.  We come in and have joint discussion with our Harmony ecosystem partners — other suppliers of VNFs and infrastructure.  Masergy is our live proof point of that because VNFs are supplied to them from ADVA, Brocade, and Fortinet — and more VNF suppliers are coming.

Customers are delighted that we can achieve their goals of agility and dynamic services at an attractive cost point.

	How clean are these NFV deployments going to be?  How much integration is required?

Certainly there’s integration work to be done, but if you design the system properly and closely adhere to VNF developer standards, it goes a lot quicker.

One of the things we’ve done is onboard a wide array of functions onto our Ensemble Orchestrator.  We made it very flexible and configurable.  So now, if we have a new VNF to onboard, it goes in very quickly.

Integration will steadily get easier as newer techniques become available.

For example, there’s a protocol called Emit Cloud that VNFs can use to get their initial configuration.  So in the case of a router, maybe it is the IP addresses of its interfaces and initial route table.  In the early VNFs, you would have to spin up a VNF and the VM and then go through the virtual serial port and configure it.  Using CloudInit there’s now a standard way to get the initial configuration.

So with greater NFV deployments, people will learn where the soft spots of these deployments are and update their software to slip in these new techniques.

	How much will NFV and SDN shake up the market of current suppliers?  For instance, what’s going to be the future role of systems integrators like Netcracker and Ericsson?

The integrators are the suppliers the operators trust.  And the operators are simply not going to rip out all of their OSS and BSS and management system.  It’s just not practical.

Instead we need to find ways to adapt and integrate new technologies like NFV orchestration and SDN controllers into those existing and evolving systems.  Now that’s a challenge, but it’s also an opportunity for the integrators — they’ve made a business out of supplying software and integration solutions.

On our side, we need to make sure that products like our Ensemble Orchestrator have flexible and adaptable northbound interfaces so it can easily tie into other systems.

	Masergy is implementing NFV at customer sites.  Why is that?  And what are the pros and cons of deploying at customer sites vs. at a network-central location?

Some of the early announced deployments of NFV have been centralized, meaning people are running VNFs on servers located in data centers.  Now that’s a good model and there are benefits to that model in terms of cost — and our Orchestrator supports a central model.

However, there are big advantages to doing virtualization at the edge.

First, you don’t need to build any infrastructure to get started.  For instance, you don’t need to roll it out to your data centers to enable virtualization.

A lot of operators have data center groups that are run separately from the telcos, so you can’t just go in and wipe servers to start spinning up virtual firewalls and routers.

One of those service providers is CenturyLink.  And they have gone out and put in infrastructure in a number of their worldwide data centers to enable virtualization, but a build out was essential before they could do anything.

So the centralized model requires an upfront investment, however, doing it at the customer’s site means you can spin up very linearly with your deployments.

Even trials are attractive at the edge because the material cost of doing demos is much lower.  You don’t need racks and stacks of equipment.  All you need are some servers and some software.

There are also certain functions you need to do at the edge or you can’t do them.  For instance, with certain forms of security, if you need to prevent unencrypted packets from leaving a building you need that on-site.

Things like QoS and WAN optimization are also best deployed at the customer site because if there’s a bottleneck due to a low-speed access link, discovering that through a central cloud is much slower.

Now that’s not to say all customers need to deploy at the edge.  You decide where virtualization functions should go by looking at what the service requires and the availability of resources.

	When I heard carrier speakers talk at the GEN15 Metro Ethernet Forum, I got the impression that virtual dynamic services in an interconnect or wholesale environment is going to take a long time to arrive.  What’s your take?

I think it’s an important area to start working on.  And I honestly believe that dynamic network-to-network interfaces will arrive sooner rather than later.

To get dynamic services, you need to solve the problem of network-to-network interfaces.  Remember, no operator can satisfy the connectivity needs of a global enterprise without using the services of another service provider.  You are forced to partner with your competitors.

Unless you’re going to nail up lots of big fat connections all over the world, you need the ability to dynamically signal connections to another operator.  And that includes things like adjusting bandwidth and QoS.

So the dynamic services will go hand-in-hand with dynamic interconnects with other operators.  And if you look at innovative operators like Masergy, they are ready to go down that path.

Some interesting work is going on at NTT already.  They are already applying SDN to their network to make services provisionable.  So we may very well see people going to fully automated interconnects relatively soon.

	Prayson, thanks for this wonderful briefing.  The Masergy case really is an eye opener and preview of what the future is going to look like.

Thanks, Dan.  Masergy is pushing full steam ahead and as you can see has fully embraced the cloud model.  They are eager to push the envelope in everything from deploying lots of VNFs and driving advanced automation to new commercial models.

Masergy’s plans are the wave of the future and we at ADVA are fully committed to supporting them.