Enabling Self-Healing, Self-Optimizing Service Provider Networks: Part 3 of 3

This last installation of our series will cover the challenge of architecting SDN controllers and how a management platform can help provide the analytics and orchestration needed to make SDN work in the WAN.

Architecting the Controllers

With all the issues surrounding controllers as discussed in part two of this series, the service provider controller deployment strategy is not very clear. There are several possible deployment scenarios. Service providers continue to pursue multi-vendor strategies and deploy multiple vendor controllers to manage different parts of the network. Heavy Reading analyst Caroline Chappell, in the white paper “Combining Management Intelligence & SDN Programmability” cites the example of setting up a VPN, where service providers “… may use one controller to program the VPN service endpoint in one geography, and another to program the other endpoint in another territory, with a third controller – perhaps OpenDaylight – programming the edge-to-edge component of the VPN.”

For better scale and reliability, service providers could segment the network into multiple controllers. However, Chappell says that by doing this “…SDN controllers risk becoming silos of management information…” She continues with a way to control this scenario:

“In order to avoid the fragmentation of network visibility and control in a multi-controller environment, {communications service providers} CSPs need a management platform that sits above and can orchestrate across all of them. This platform services as the locus for the common management intelligence needed to drive SDN programmability in different network domains, while maintaining a consistent, hyperfast experience for enterprise users.”

WAN SDN Architecture

Taking this idea of a management platform further, the future of SDN depends on the applications that can manage a complete life cycle of complex workflows – visibility, assurance and provisioning – in one place as opposed to the use of multiple tools. Although vendor announcements focus on SDN applications such as bandwidth calendaring, in reality it is not possible to go to market with the assumption that a network supports only SDN-enabled equipment.

In order to support hybrid networks, vendors need to bundle additional discovery and programmability capabilities. This is driving vendors to develop an infrastructure layer to support application requirements and isolate them from southbound programming modules. The goal is to package the vendor-specific, value-add modules to augment controllers and keep the controller code open. For instance, a vendor infrastructure might bundle an orchestration module that could seamlessly work with legacy provisioning tools and SDN controllers to provision hybrid enabled networks.

Three-Tier SDN Infrastructure with SDN Management and Orchestration

Figure 1: Three-Tier SDN Infrastructure with SDN Management and Orchestration

Automation is a key tenet of such agile networks, and SDN helps solve a number of automation challenges. However, as discussed in part one, without analytics automation is relegated to basic scripting and will not achieve the true potential of self-healing, self-optimizing networks. This goal can only be achieved with a closed-loop management lifecycle of real-time network telemetry collection, analytics and optimization, and automated provisioning of network services.

Real-time analytics

Figure 2: Real-time analytics

A management platform – such as the Packet Design SDN Management and Orchestration Platform – powers the future of self-healing and self-optimizing networks by providing superior service visibility and comprehensive real-time network analytics. With the help of the platform, service providers can easily develop complex workflow automation of IP/MPLS management. They can also use network behavior analytics to make investment decisions around network optimization and network expansion to match service growth.

To solve the multi-controller issue, the Packet Design platform is open. Customers can choose to use any available off-the-shelf controller. Our goal is to support as many controllers as possible and enable seamless service orchestration across multiple controllers. The platform provides a policy engine to store application and service policies. These policies are translated into underlying controller policies during provisioning.

To prevail in the wave of SDN disruption, service providers must leverage decades of experience building massive, reliable WAN networks and develop solid automation. They need to look beyond SDN and NFV and start investing in broader automation to leapfrog the competition.