Steve Harriman – Senior Vice President Marketing at Packet Design talks about Packet Design’s solutions and products and shares his knowledge about the self-driving network in this weeks’ thought-leaders on emerging technologies.

Packet Design was founded in 2003 and is transforming the way network operators plan, deploy, and assure services to help them accelerate time to revenue, decrease churn, and contain costs. Modern IP/MPLS networks call for next-gen analytics and orchestration software but most management tools fall short. With real-time telemetry for millisecond by millisecond visibility into service paths, historical and predictive analytics, and automated service provisioning, the Explorer Suite is future-proof management software that is used in hundreds of the world’s largest, most complex networks.

Keep reading to find out more about Packet Design, the self-driving network, Steve’s view on the biggest innovators in the industry and the future of the self-driving network.

What do you specifically focus on in your role as SVP Marketing?

As with any marketer, my job is to understand what the market wants and align the company’s offerings with those needs. At Packet Design, I work closely with the head of product management to ensure our products are positioned correctly and that we clearly communicate their business and technical value to our target audiences. A second area of focus is to ensure Packet Design is well positioned for the wave of change happening in our industry – network transformation, and all that it entails, including software defined networking and virtualization.

Why did you join Packed Design?
I joined when the company was acquired by the PE firm, Lone Rock Technology Group. I was part of the acquiring entity, along with several other members of the executive team. This is a team of folks that I respect and enjoy working with. We have worked together for many years at other successful companies.

Lone Rock acquired the company because of its tremendous potential. The engineering talent here is world class, and the Packet Design technology solves really hard problems, uniquely. In 2013 when we did the acquisition, software defined networking (SDN) was in its infancy, but we saw a great opportunity to leverage and build on the technology to enable adaptive, analytics-driven networks. These are sometimes called self-driving networks, a term coined by Dr. Kireeti Kompella at Juniper.

What does Packet Design focus on?

Packet Design provides management software for wide area networks. Our Explorer suite of products provides visibility into the network’s control plane – so network engineering, planning and operations teams can understand in real time or at any point in time how service traffic is routed across the network, and how routing behaviour affects service delivery.  Our software also provides what-if modelling, so the impact of network changes, failures or new workloads can be predicted ahead of time.

This is quite different from traditional tools that poll network devices every few minutes to collect utilization and performance metrics, or those that periodically retrieve configuration data so they can map the network’s topology. These are fine tools, but they cannot supply the real-time intelligence needed for self-driving wide area networks. For that, you need real-time routing, traffic and performance telemetry.

I think it’s helpful to use a simple analogy. Think of Google Waze compared with the paper road maps we used to rely on. A road map would show us the most direct route from point A to point B, but it could not tell us which route would be the quickest because it provided no information about road construction, traffic accidents, police activity, anomalous traffic caused by a big sporting event, etc. Google Waze, on the other hand, provides guidance on which is the optimum route based on real-time inputs. That’s what Packet Design does for your IP/MPLS network.

What is a self-driving network?

A self-driving network is one that adapts to changing conditions without human intervention. It automatically adjusts to demands on the fly, detects threats to service delivery and makes necessary adjustments, and self-optimizes to use network resources most efficiently. The industry is at the beginning of this journey and we will see these autonomous networks evolve over many years.

For example, a customer in Southeast Asia is using our technology to automate recovery from network failures to avoid service interruptions. It’s not uncommon for subterranean or subsea links to be compromised. When this happens, our software analyses the current network state, predicts loads based on historical traffic patterns, link utilization and capacity, and uses various algorithms to calculate the optimum alternate path for the affected traffic. It then sends the network configuration changes to the OpenDaylight SDN controller for automated provisioning. All of this happens in just a few minutes versus what would normally take several hours or days.

How did the self-driving network originate?

It is a natural evolution enabled by real-time telemetry, virtualization and other technology advances, as well as industry standards and open source initiatives. We have already seen the emergence of software defined data centres, where compute resources are intelligently provisioned and de-provisioned on the fly. The wide area network is the next frontier, although I would say it’s a much harder problem.

What are the benefits of a self-driving network?

Improved service delivery, greater business agility, and lower costs.

Today’s networks are complex and support a wide variety of services, including corporate connectivity, business applications, video streaming, online gaming, data replication, and so on. Each has its specific quality of service requirements, and the workload mix can vary dramatically depending on the time of day or day of week. Designing, configuring and maintaining networks to support this mix is complex and labour intensive. Adding or changing services can take days. If the changes can be automated, based on network analytics and business policy, service activation time (and therefore, time to revenue) can be shrunk dramatically.

Similarly, if threats to service delivery are detected and remediated automatically (as illustrated by the Southeast Asian use case I mentioned earlier), loss of service revenue or SLA penalties, in the case of a service provider, can be reduced. Customer satisfaction is preserved, churn reduced, and brand-damaging incidents avoided.

What are the weaknesses of a self-driving network?

I’m not sure about weaknesses, but the challenges are technological and cultural.

The former will be solved over the coming years. Large packet and optical networks are complex, multi-vendor infrastructures. For a self-driving network to function, all network components and analytics and orchestration software must inter-operate; hence the need for standard protocols and open APIs. As always, there will be a tension between the need for vendors to maintain competitive advantage via proprietary technology and the need to support open standards. I’m confident this will be resolved – things are a lot different today than twenty years ago. The market demands it, plus there are more open source alternatives.

There could be some cultural resistance to automation: What will be the impact on jobs? If history is any guide, the jobs will morph but they won’t go away. We have more automation in IT today than at any time in the past, and the job market is as buoyant as ever.

Who are the biggest innovators when it comes to the self-driving network?

There is a lot of exciting work going on in this area and I don’t see one innovation leader. Most of the SDN and virtualization deployments we see are multi-vendor ecosystems. Not one vendor can do it all, despite what their press releases say. The open source community is doing great work: examples include the OpenDaylight (ODL), ONOS and ONAP. Packet Design is a contributor to ODL.

Most of the real innovation is in software, like machine learning and AI-based analytics, and this is where independent software companies run circles around the network equipment manufacturers. I think Packet Design’s SDN platform and applications are right up there when it comes to innovation.

Where do you see the future of the self-driving network going?

The need for adaptive networks is just going to increase. With so many demands, particularly IOT, being placed on networks, intelligent automation is required. Network services will be stood up and taken down without human intervention, in the same way virtual machines and platform instances are provisioned and deprovisioned in the cloud. The network’s physical packet and optical layers will provide a flexible underlay that supports the provisioning and deprovisioning of virtualized overlay services, driven by real-time analytics and business policy.

What skills do Network Engineers need to stay relevant in the Networking industry?

Automation will reduce the need for humans to manage some day-to-day tasks, like monitoring and reconfiguring routers. But designing, planning, building and maintaining network infrastructure will still be necessary. Engineers should learn scripting/high-level programming languages, like Python, as these form the glue in multi-vendor ecosystems. And, as with any technical job, understanding how to apply technology to address business goals is always good for one’s career.