This post is also available in: Japanese
Patrick Ostiguy founded Accedian Networks in 2004 and serves as President, Chief Executive Officer and sits on the Board of Directors. He has nearly two decades of telecom industry experience and is responsible for filing several patent applications.
SDNCentral: Can you please provide us with an overview about Accedian, and the problems you solve?
Patrick: ”In a nutshell, we are the leader in performance-assured networking, which we consider to be the perfect blend between packet networking and performance assurance. At the base of our DNA is an architecture that allows for doing these two things in congruence. This is something we have today deployed in more than 150,000 locations around the world. We do this primarily for mobile backhaul, and mostly in North America and Western Europe.
Another area of involvement has been in business services, more specifically, performance-assured Ethernet as a means of connecting enterprises to cloud-based services.”
SDNCentral: What markets do you operate in, and what is the role of SDN in these markets?
Patrick: “We provide solutions for mobile operators, service providers and enterprises. In the case of mobile backhaul specifically, it is still very early for SDN to have a really wide presence, but the interest is there. We are starting to see cases where SDN is being used for pilot projects.
Mobile operators have adopted an automation technology called self-organizing networks (SON). SON is designed to make configuring, managing, and optimizing mobile radio access networks simpler and faster. Operators have come to expect the same kind of mechanism when it comes to backhaul and are looking for ways this can be achieved. The heterogeneous backhaul network is in dire need for the intelligence provided by SDN and the self-optimization capabilities provided by SDN technology.
Without it, the pain points are gross underutilization of backhaul and the inability to plan for bandwidth upgrades altogether. At the end of the day, customer experience suffers dramatically from such a leaky network of pipes that the providing network elements are totally blind to.”
SDNCentral: What form do your Performance Assurance solutions take?
Patrick: ”Essentially they take the form of what we call performance-assured elements, which are instances that are installed, implanted, or disseminated in the networks in the form of things like software agents, miniaturized network devices and performance-aware network elements. They harvest performance assurance data, which they then convey up to performance assurance aggregation nodes higher in the network. The nodes collect the data, crunch it, and push the aggregated data toward overhanging applications that can then present it to reporting portals or actually generate portals.
We specialize in multivendor environments by providing standards-based solutions. In terms of network state information, traditional “still” snapshots from packet counters work fine in data centers, but they are not sufficient across complex networks or for telco SDN. Accedian’s goal is to extend the ownership of performance assurance to multiple vendors and really create the ability for the operators to have an open environment where performance is assured. We provide complete, end-to-end, in-motion pictures of transport performance with visibility across multiple mediums, multiple technologies, multiple carriers, and multiple domains.”
SDNCentral: What do you see as the role of performance assurance in an SDN-driven network?
Patrick: ”SDN was born in data centers, and it was built on the premise that the interconnection between network elements would be robust. As SDN is just now expanding outside of data centers into the WAN or telco SDN, there’s inherently a lack of the kind of performance assurance capabilities that telcos have come to expect from their network. As you start to cross multiple domains, multiple carriers, multiple technologies, multiple mediums, you get into scenarios where performance is impaired in a variety of ways that is far from what you would see in a data center. The initial instances of SDN for telco are ill-equipped to have the visibility that allows them to have the same levels of optimization as what they’ve been doing in the data centers.
In order to address this, we’ve created a suite that’s called Network State+, which is a suite of parameters that provides visibility across multiple media, technology, carriers and domains. It gives added insight to account for things like latency jitter and packet loss, and it provides visibility of available bandwidth across links and on a hop-by-hop basis.
Providing the available bandwidth in real time is absolutely mandatory if you want to optimize the network dynamically. There is an enormous amount of stranded bandwidth currently underutilized on a hop-by-hop basis, whether it’s bandwidth stranded for redundancy purposes, because of over-engineering, or because of lack of visibility or lack of traffic engineering tools. The hop-by-hop capability to see this allows a higher degree of optimization.”
SDNCentral: Can you describe your view of an SDN architecture with a performance assurance component?
Patrick: ”The current state of things when it comes to mobile backhaul heterogeneous networks is essentially a network made up of leaky pipes. At the top you have a router blindly sending data all the way down to recipients, which are small cells and then end users. These network elements such as switch routers from the usual OEM vendors have very little if any visibility of the performance of individual links they’re trying to throw traffic down to.
As a result, these network elements have been manually set to a fixed type of setting like CIR or EIR. The available bandwidth varies across pipes whether because of varying atmospheric conditions, bandwidth sharing between operators, misalignment of various antennas, or interference at the spectrum level. You get a dynamic picture when it comes to these individual links for which the overhanging network element has no visibility. Network elements try to send data down by trial and error, causing a lot of spillover and wasted bandwidth, especially at the top of that particular type of architecture.
The new proposed way of doing this is to fundamentally add that visibility. Give that visibility to the network element at the top of the tree by connecting to a control layer that is SDN enabled. This application layer is performance aware because of Network State+, our toolset that provides visibility of the network state in real time across layers 2-4. The application layer is then connected to the control plane by the Network State+ API.
Of course, the primary data has to come from somewhere. The data, the way it is proposed, comes from the actual network elements themselves, in the form of performance assurance software agents that have been embedded in a standard way or a proprietary way. This is conveyed to a software application called a performance assurance actuator, which sits in that specific network element and can then aggregate, collate, collect and forward the aggregated performance assurance data to the Network State+ application.
Fundamentally, the idea is to make the SDN controller performance aware, give visibility of the router or switch router, and provide the ability to dynamically react to these changes.”
SDNCentral: What is SkyLIGHT? What does it provide to customers?
Patrick: ”SkyLIGHT is Accedian’s solutions architecture for performance assurance. It is comprised of three types of elements. First, you have in the network itself performance assurance elements. These are very generic but can range from software agents that are running in an embedded fashion in any kind of network element you can think of, whether it’s radio nodes or switch routers. It can also take the form of miniaturized small form factor pluggables that have the same role as the software agents, but where embedding it has not been possible for time or capacity reasons.
The second angle of SkyLIGHT’s solutions architecture is performance assurance nodes. Those are the actuators that in a physical format would have the ability to collect large amounts of primary performance assurance data. These can also be virtualized — and this is key because as networks start to migrate toward NFV, we see these kinds of functions are becoming more and more virtualized. Networks will take the mediation, collection and capability to aggregate performance assurance data and virtualize them as much as possible.
Finally, the overhanging application layer is what takes this performance assurance data and translates and shares it with SDN and SON controllers. It takes the performance assurance data that has been generated by the element, that has been collated and aggregated by performance assurance nodes, and allows it to be shared with either other applications or with SDN and SON controllers. This architecture provides operators with the ability to take the promise of SDN and apply it to mobile backhaul networks by providing the level of insight and visibility that is currently lacking.”
SDNCentral: What is your timeframe for delivering SkyLIGHT to production deployment?
Patrick: ”Different aspects of the architecture are virtualized to different degrees, and we’ve already started to deliver some solutions to our customers. The software agent piece is something that is already available, as are the actual implants, or small form factor pluggables. We also have actuators providing a virtual NID functionality, removing the need for dedicated NIDs.
Now we’re looking at taking this one step further in terms of virtualization and taking the additional step of providing the network assurance application to NFV and SDN environments. We are targeting to have general availability for the end of this year. We are working with several partners to make it happen.”
SDNCentral: What are natural partners for your SkyLIGHT framework?
Patrick: ”We currently are embedded into Blue Planet, an SDN system from Cyan that consists of an open SDN platform and apps. We’re also working with several providers of SDN controllers and certainly are looking at OpenDaylight as well.
We can’t name all our partnerships publicly, but the partnerships we do have in place with the various base station vendors cover more than 80 percent of the base stations in the market. This makes our solution not only standards-based, but fully interoperable by design, with 80 percent of the base stations out there.
In terms of other partners, as Network State+ will be sharing network performance assurance data within the application layer, we also are working on partnerships with other application layer providers.”
SDNCentral: What would be your advice to customers in your market looking at SDN? What should they be thinking about?
Patrick: ”The most important advice is to try as much to extend ownership of performance all the way to multivendor base stations. I think having an independent application provider really guarantees that you keep this ability. There is a multiplicity of base station vendors out there. This increases even more with the small cells and heterogeneous networks being deployed. And for this, you definitely need a multivendor approach and an open approach as opposed to having the fox watching the hen house. An independent point of view is more and more required when you have such a zoo of varying performance and multiplicity of vendors.
The other piece of advice I would give is not to underestimate the level of complexity that this causes. Go for a solution that is simple and streamlined. Also, do not underestimate the speed with which LTE is filling up these microwavelinks. With LTE, you open the floodgates and basically have filled up the pipes instantaneously. That is something that certainly requires optimization right off the bat.
The good old “throwing more bandwidth at the problem,” or “let’s assume infinite bandwidth” does not apply in this case because there’s an imbalance that’s been created. The speeds that are enabled by LTE fill up so quickly the mobile backhaul links are not optimized. Optimization was one of the main drivers for SDN, so make use of it.”
SDNCentral: What key milestones for performance-assured SDN do you see in the next 18 months?
Patrick: ”I would say one of the biggest milestones will be the adoption of an SDN controller plane by a mobile operator in a real live environment, as opposed to a pilot trial. Several pilot trials are out there – I can count four off the top of my head. But to go full, network-wide deployment will be an important milestone to achieve. We sense this could happen first in environments where optimization is a dire need, mainly where you have heterogeneous networks like small cells and macro cells and where you have wireless being used as a means of backhaul. We plan to roll out SkyLIGHT integrated with a control application specifically for optimizing mobile backhaul.”
SDNCentral: Patrick, it’s been a pleasure speaking with you. Thanks for sharing your time!
Check out more Interviews on SDNCentral:
- A Chat with Thomas Nadeau and Ken Gray, Authors of O’Reilly’s SDN Book
- Stateless Networks, the ‘Gateway Drug’ to SDN? A Chat with Kelly Wanser, CEO
- SDDC13 Speaker Viewpoint: Software Defined Data Center — A Regulated Enterprise Perspective, by Andy Brown, Group CTO, UBS
- Software-Defined Data Center Symposium Q&A: Greg Ferro of Packet Pushers and Ethereal Mind
- SDDC Q&A: Ivan Pepelnjak and Brent Salisbury