For this post I met up with DartPoints CTO Satya Baddipudi to discuss where best to locate virtual functions, when NFV will hit the mainstream and how a software approach to service chaining can replace the “conga line” of devices.
Prayson: Satya, could you start off by telling me a little about your company's role in the NFV transformation?
Satya: At DartPoints we had been taking a cautiously optimistic view of the NFV technology, and now we’re moving to be one of the early adopters. Most of our deployments are greenfield, which is a little different from the other ISPs and CSPs. This enables us to be receptive to new technologies and innovations, wherever it makes sense within our own risk acceptance.
Today we offer network and security services that are based on hardware appliances such as firewalls and routers. Going forward, we expect some of them to be software-based VNFs, especially the ones close to our customers. This enables us to deploy services rapidly and cost effectively.
The other important thing is scalability and our operational simplicity. By the end of 2016 we expect to be deployed in 10 or more sites, so NFV will be very important for our operations.
Prayson: You mentioned using NFV for applications closer to the customer. Are you looking at both hosting VNFs in your data centers as well as pushing them out to the customer sites?
Satya: Yes. In networking terms we classify our sites into two tiers, one as core sites and the other as edge or access sites.
We may or may not offer our co-location services at every access site. It would be more about focusing on getting them connected to our network services, which could be just a VNF deployed at the edge sites. Services such as colocation and cloud can be offered from the core sites.
Prayson: That’s a very realistic approach – putting in the functions where they need to go based on the needs of the service. You mentioned that you’re an early adopter. What obstacles to NFV have you seen and are you working to address?
Satya: As expected, most of these obstacles are not unique to us. They come from operations and support systems that have to work with a newer technology. If we take a step back the way we envision using these services fundamentally remains similar to what our customers are used to for today’s services such as managed firewall or managed Internet access. They won’t see much of a difference. It’s just that the underlying enabling technology will be changing from hardware to software.
Prayson: Do you see that customers will benefit in terms of the dynamism of the services?
Satya: Yes, definitely. For example, if they were to go and procure two firewalls and two routers, it would take some time to get the service up. With NFV it only takes a few minutes to create the service by spinning up a VNF.
Prayson: Do you feel that you have the staff and expertise to move from appliance-based functions to software virtualized network functions?
Satya: Absolutely. It’s going to take training in some areas, but in others we can leverage our existing expertise. For example, we partner with technology vendors such as Fortinet, Cisco and others for firewall appliances. When we move to a VNF version of that firewall, the guy who is operating and provisioning it won’t see much of a difference, because it has the same user interface.
But the other support teams need to be educated and trained on the underlying virtualization technology.
Prayson: As you mentioned, you’re on the leading edge here of using NFV. When do you see other people using NFV in a more mainstream fashion, such as your competitors and people in other industries?
Satya: I believe that it is coming but it’s going to be difficult to predict when it will hit the Tier 1 ISPs and CSPs at full scale. We know that most of them are working toward their own solutions. Being big companies it will take some time for them to just get through the project management cycles on the way to the rollout.
I think it may not be the technology by itself causing the delays. I think it’s more of the support training, operations, billing, etc., that need to fall into place. If I were to take a guess, I would say it will be early to mid-2017 when more people will get comfortable with NFV.
And along with that, I also expect that some level of consolidation and standardization needs to happen before people move forward. When there is a new release of the Windows operating system, people like me always wait for service pack 1.
Prayson: (laughs) In the meantime, is there an opportunity for companies like DartPoints to go out and win some business based on being able to offer your customers more dynamic services?
Satya: Yes. I think we’re seeing an immediate benefit. These dynamic services are not yet available with other ISPs and CSPs. They already have established processes and procedures in place for the current services and it will take a longer time for them to change.
Prayson: What is the one thing that’s been most surprising or unexpected as you’ve moved down the path toward NFV?
Satya: I’ve always had my eyes on SDN and NFV as technologies. About four years ago, I started searching for a technology as a stable software-based solution with service-chaining capabilities for remote offices. At one of my previous companies we called it a “conga line,” where at any remote office we had this lineup of devices: a router, a firewall, a WAN optimizer and an IPS appliance.
This lineup was difficult to install and maintain. It would be nice to have something that we could manage remotely with a software approach. The technology at that time was not straightforward, too expensive or non-existent.
So coming back to your question: something that surprised me was how people perceived NFV as a completely new technology. If you look at the underlying technology, it’s virtualization with service chain abilities.
Virtualization by itself has been a mature technology for a while. But what we are now seeing is this value being brought into the network services domain.
Prayson: I like your metaphor of the conga line. The image keeps going through my mind. It’s very relevant and I will be using it in the future. Thank you for that.
Prayson: You said virtualization is not new and I agree. One of the drivers for NFV was that the service providers wanted to take some of the benefits they were getting with cloud services and drive them into the network. One of those benefits was new commercialization models.
You mentioned your use of network appliances. You buy the appliance and it costs what it costs. Operators would like to go to a model where they buy a function and they pay for what they use. They only pay when they use it, and what they pay is related to revenue generated. Are new commercial models a benefit that you’re trying to achieve in discussions with your suppliers?
Satya: Absolutely. The important value for us comes from the fact that I’m not buying router or firewall appliance. Instead I’m buying commodity servers and then I’m buying software licenses for what I need.
The big value for us comes from the ability to quickly spin up these VMs and then have the ability to provide the services in a much quicker fashion. I can spin up these VMs as needed. I don’t need to keep hardware appliances in inventory expecting that the customers will be coming in the future.
Prayson: You are a believer in NFV orchestration. What are the benefits of going with a full-featured orchestrator?
Satya: NFV is straightforward for most businesses. They buy the appliances and VNFs and spin them up one at a time. It’s within a limited support domain.
But as a service provider, we go out and provide these services to external customers, and we are bound by strict SLAs. A centralized service orchestrator is critical from the operations and support standpoint. Having the ability to manage multiple sites from the same platform is going to be extremely important for us.
Prayson: That’s consistent with how we portray the benefits of orchestration. Automation and assurance of the full life cycle along with management and scalability are all very important points.
What’s next for DartPoints?
Satya: One of our advantages is that we are a distributed micro data center company and we are next door to a lot of enterprise businesses. Our focus has been micro data centers, which by our definition are below 500 kilowatts and are located in commercial buildings and business parks. This enables us to function both as a data center provider and also as a CSP.
A lot of these businesses have sub-100Mbit/s and few at 1Gbit/s internet connections, and they’ll buy routers and firewalls and implement their own service. We can now work with them to provide a self-managed or co-managed service network as a true utility service.
People always talk about the bandwidth as a commodity but I don’t think that’s ever been realized in its entirety. So I think our approach enables us to provide a network utility service, similar to the way we pay access water or power on a usage basis.
So we’re looking at a new way of delivering services to customers. I think it remains to be seen whether customers are ready to accept this kind of true utility service.
Prayson: One thing that’s clear to me is for that network-as-a-service model to work, it needs to be based on software implementations and automation as opposed to physical boxes and cables. Would you agree with that?
Satya: Definitely, and I think there’s a place for both software and appliances. Today can I use a router VNF to go peer with an ISP and learn 500,000 routes? Probably not. I’m going to buy a Cisco or Juniper router to do that. But on the other hand, I’m not going to buy the same thing when I’m just connecting to a couple of ISPs on a 100Mbit/s connection.
I think every technology has its own place. But NFV opens the door to new way of looking at things.
Prayson: Thank you, Satya. We appreciate hearing about your innovative business model and how NFV orchestration is a key enabler for DartPoints’ growth in 2016.
The Real CTOs of NFV is a series of conversations where I talk with CTOs of leading service providers and suppliers to find out what’s happening in the NFV space.