Accedian Networks provides performance assurance for mobile operators, service provider networks, and data center connectivity. We talked with Accedian President and CEO Patrick Ostiguy about how the company provides network visibility to help optimize performance.
SDNCentral: With the growth in both mobile and enterprise connectivity and all the excitement around SDN and NFV, how has this affected your customer base over the past couple of years?
Ostiguy: Historically, we’ve seen performance assurance needs primarily from mobile operators and service providers. However, with the growth in data-center-to-enterprise connectivity, and increased flexibility and dynamism in these networks from SDN and NFV initiatives, we’re seeing a large jump in performance assurance needs from the enterprise. Today, half our business comes from mobile carriers, and the other half is data-center-to-enterprise connectivity. The enterprise side of the business has grown tremendously.
What are some examples of enterprise use cases driving this growth?
For enterprise-to-data center connectivity, there is a mix of performance assurance demand from enterprises directly, service providers connecting businesses to data centers, and cloud providers with their own networks who extend to customers through local provider access. Unlike historical hub-and-spoke, remote-branches-to-headquarters connectivity, we’re seeing a shift to mesh connections to data centers in multiple locations, both private clouds and public clouds.
To illustrate the scale of this growth, one very large enterprise customer is linking its remote branches to its main data centers. That’s happening globally, spanning more than 14,000 sites. In cases like this, a lot of connectivity that used to be in place has to change and adapt to new topologies, new services, bandwidth needs, requirements, and technical specs.
What kind of evolution have you seen in customer needs on the enterprise side?
Frankly, the enterprise never fully embraced traditional metro Ethernet, or MEF (Metro Ethernet Forum)-type service specs. We’re seeing a shift towards a lot more emphasis on bigger pipes, faster bandwidth, and more requirements around leveraging bandwidth for data center applications than Carrier Ethernet service attributes.
I would say performance assurance requirements are going from basic Layer 2-3 toward TCP and the application layer. People want performance from headquarters to remote branches to be as good as if they were connected directly, even though it’s going through a data center.
And on the mobile operator side of the business, what changes have you seen?
Certainly we’re seeing much more traction on our virtualized portfolio, but we’re still doing well with our traditional hardware NIDs [network interface devices] for traditional demarcation. We’ve been changing the model the last few years to adapt to the rise of virtualization and NFV. We’ve been reducing the amount of hardcoded elements within the units’ FPGA and embedded CPU, and migrating these functions into software that can be run in a virtualized environment. In situations where we can cost-effectively use an FPGA without unnecessarily tying ourselves to hardware, we’ll still do so. Nevertheless, we’re moving more of our performance assurance application into virtual machines that can reside in a data center — this NFV approach allows us to scale our performance assurance functions to tremendous scale.
By limiting the amount we do in hardware, we were able to minimize our NID into an SFP form factor, creating a breakthrough cost model for network instrumentation that’s really disruptive compared to traditional solutions.
Because the FPGA is focused only on what can’t be done in software or merchant silicon, it’s allowed us to do more. Our Nano module SFPs can now do full-blown service activation testing at full line-rate. They have the same kind of firepower as you would find in a portable test set priced around $15,000.
By tying this data-gathering framework to our NFV approach, we have centralized, virtualized, scalable intelligence, but with distributed test traffic generation and termination — much like how SDN centralizes the control plane and simplifies packet-forwarding functions in white boxes. We put the minimum hardware required at the edge and virtualize the rest. That’s our take on NFV — the most efficient blend between hardware, software, and virtualization.
How about vNIDs? Are there situations where there is no physical access and vNIDs are deployed?
Yes. Today, at large Tier 1 service providers, we’re having great success with the centralized/virtualized intelligence model, which is the vNID model combined with our Accedian SkyLIGHT API.
Our V-NID platform supports a large number of OA&M [operations, administration and management] and performance monitoring functions supported by standards-based network elements, base stations, small cells and so on.
In these cases, the network itself becomes the test endpoint, but there are often blind spots. It could be because the native functions only support Layer 2 and the customer needs Layer 3 visibility, or because standards are not implemented completely by third-party hardware, or because standards are not supported at all — as in the case of low-cost endpoints like small cells. In these cases the Nano Module SFPs complete the required instrumentation coverage for V-NID deployments.
The combination of vNIDs, Nano smart SFPs, and our traditional NIDs gives us a solution that spans all the critical measurement points for cloud-based applications, from the VM to the data center switches, to the routers, the WAN devices, through mobile networks, all the way to base stations and mobile devices.
What kind of feedback are you getting from your customers about NFV and VNFs?
They are seeking disruptive economic savings happening at the customer premises. You’re really changing the model if you can have great hardware manageability in a virtualized, centralized, and scalable fashion and keep customer premises equipment to the bare minimum.
Whether it’s the deployment of small cells or data center connectivity, cost is extremely sensitive. That’s why we are seeing increased enterprise connectivity over virtual networks overlaying the open Internet — OTT WANs. If you’re going to go over the top with the access piece, service assurance becomes even more critical for CIOs to meet their internal SLAs. With virtualized instrumentation you can have a real-time view that allows performance to be optimized with feedback as they tweak their access. This allows OTT WANs to be as good as or even better than MEF- or Ethernet 2.0-type schemes.
You’re going to disrupt price with this model, because now you’re not worrying about customer premises devices that have to be maintained, guaranteed, or serviced. We’re seeing a tremendous amount of interest in this with data center, enterprise connectivity, mobile backhaul, and small cell. And as a result, customers are putting more NIDs in more locations to provide them with the necessary visibility and control over network connectivity at these points.
With that many measurement points, how do you keep track of data coming in to your centralized intelligence? For instance, if you have 14,000 locations sending fine-grained information, how do make sense of that?
We’ve created a concept called Network State+, where we’ve taken the raw metrics for basic delay, jitter, packet loss, etc., and created an extra set of metrics to provide more meaning: one-way measurements, derived QoE scores, statistical views, very granular and precise KPIs.
We created metrics that we make available to third-party applications that have means of being the eyes and ears of optimization. The idea is to work more closely with SDN vendors, where we basically become the network instrumentation layer.
Picture a three layer cake: the bottom layer is the network elements, which all have standards-based OAM and PM at Layers 2 and 3. In the middle you have Accedian solutions, which is a multivendor network instrumentation layer, and at the top you have the BSS, OSS, NMS, or SDN system.
We act southbound to collect precise network QoS from the network itself using virtualized instrumentation, and we create more actionable information that we feed northbound to the higher-layer systems, which then use the information to make educated decisions about optimization.
How has your product line or mission evolved over the past year or so? What can we expect from Accedian in the future?
As virtualization and software-defined networking start to take a more predominant role, we’ve shifted from being a demarcation player and more of a player in the network instrumentation layer — a solution, instead of point-products. We’re taking a more important and strategic role.
Right now, we’re working on combining physical and virtual path assurance. Providing that combined picture is going to be crucial to the initiatives of large Tier-1s when it comes to their network 2.0 evolutions.
This allows you to get a picture of performance that reaches from the IT closet through the network, all the way to the individual virtual machines. It allows us to ultimately monitor both the virtual and physical paths that are being utilized, and correlate the two.
We’re seeing a lot of changes happening in the industry. It’s really quite an exciting time for us. I think we’re in the right place to help the SDN shift accelerate.