Light the way to optical SDN reality with test & lab automation

June 25, 2013
Software-defined networks (SDNs) offer the opportunity to make optical networks more agile and efficient in delivering bandwidth to applications and services in data center, metro, and transport networks. SDN’s bright potential is matched by the shadowy challenge of evolving existing networks to not just a new protocol but a new paradigm. To ensure that SDNs are deployed smoothly, it’s important to evolve critical business processes such as equipment and service testing to be as agile and efficient as SDNs promise to be.

Software-defined networks (SDNs) offer the opportunity to make optical networks more agile and efficient in delivering bandwidth to applications and services in data center, metro, and transport networks. SDN’s bright potential is matched by the shadowy challenge of evolving existing networks to not just a new protocol but a new paradigm. To ensure that SDNs are deployed smoothly, it’s important to evolve critical business processes such as equipment and service testing to be as agile and efficient as SDNs promise to be.

The benefits of SDNs to optical networks
SDNs first came about and are still primarily deployed in the academic community. They have also rapidly become the “next big thing” in networking writ large. SDNs offer a separation of control plane (path or route computation) from data plane (packet or optical switching) functionality. By doing so, expensive networking hardware that currently packages both control and data plane functionality can be commoditized into simpler data plane switches and standard servers running control plane software. In addition, centralized SDN controllers can more effectively offer interfaces by which applications, services, and network management systems (NMSs) can dynamically request service paths, making the network more responsive to evolving business needs and requirements. Finally, SDN controllers offer a single point of change for network elements—since most product changes don’t occur in the silicon-driven forwarding plane, but in the control and management software features and interfaces.

While SDNs’ initial focus was on data center and IP network environments, they also hold real benefits for optical networks in both transport and data center environments.

Carriers have long desired to make transport networks easier to provision, friendlier to exploding volumes of packet data, and more agile in terms of bringing new services to market to combat the erosion of margins from over-the-top (OTT) services. Prominent initiatives in this direction are GMPLS from the IETF and MPLS-TP as a joint project of the IETF and ITU. The challenge for these initiatives is that IP/MPLS network control planes are inherently distributed, so the adoption of these standards to create a unified control plane for optical/transport and IP/MPLS packet networks has been challenging.

SDNs promise to help the evolution to agile optical transport networks and aid the integration of transport and packet networks:

  • The SDN paradigm involves centralized controllers, which fits very well with optical transport network control and management plane paradigms, which are more centralized than IP/MPLS networks are. With SDN controllers as the single point of interface for NMSs and external applications, optical/transport and IP/MPLS paradigms have a more ideal unified control plane.
  • A holy grail of network service agility is the ability to rapidly provision network bandwidth to meet changing traffic requirements. SDNs provide northbound APIs to applications and NMSs that can dynamically set up service paths across the network. One can envision a scenario in which a new viral mobile application or a major news event that dramatically shapes traffic patterns would be matched by the network provisioning system to the bandwidth necessary to deliver that traffic effectively. Another use case is where Infrastructure as a Service (IaaS) services burst virtualized data center resources and require increased inter-data-center bandwidth, which could be responsively provisioned to maintain an agreed upon class of service (CoS) service-level agreement (SLA). Since major bandwidth shifts may require provisioning of multiple network layers, the extension of SDN to optical transport would assist in making these scenarios reality.

Data centers can also benefit from optical SDNs. Just as rapid changes in traffic create the need to offer dynamic service topologies consisting of packet and transport layer bandwidth, data center networks also have a need to rapidly provision massive amounts of bandwidth between server and storage elements to enable effective service delivery. One of the promises of optical SDN is to provide a way to converge ultra-high-bandwidth optical switching and traditional packet switching into a single, unified data center fabric that can optimally respond to bandwidth needs of varying scale.

Why testing needs to become agile too
A reality check on this bright future is that SDNs won’t displace legacy architectures. So to deploy SDNs requires careful planning and testing to ensure that they will deliver on their promises in real networks. Inserting SDN functionality into an operational network requires rigorous testing of the full matrix of existing and new network hardware and software components. This includes testing the northbound API calls between multiple controllers, NMSs, and applications and the SDN protocol messages to a variety of transport and packet switches. Testing also needs to factor network load, resilience, and performance factors.

A key investment that engineering groups can make in preparation for this complex network evolution is test and lab automation. The reason is that while SDNs promise greater agility for data center and packet/transport networks, test lab operations are still stuck in a highly manual operational model. It is not uncommon today for testers to take days to physically connect and configure a single test topology. Due to the time it takes to set up tests, equipment sharing is typically low, which means that organizations must either make massive, recurring investments in lab equipment or tests must wait. Tests are also most often manually run, since the vast majority of network testers aren’t programmers. Together, these “hand-driven” processes make the testing phase of new products and services painfully slow.

What’s needed is an automated, software-driven approach to test labs, where software is used to:

  • Manage all the physical and virtualized network elements
  • Easily find devices and create test network topologies of any level of complexity
  • Schedule resources
  • Configure and physically connect equipment
  • Save and restore full test topologies
  • Drive test automation development without the need for programming skills.

SDNs promise greater agility for transport and data center optical networks. Making test and lab operations more agile through automation software can help ensure that the promise will become reality.

Alex Henthorn-Iwane joined QualiSystems in 2013 as vice-president of marketing, and is responsible for worldwide marketing and public relations.