Light Reading’s recent webcast on Managing & Orchestrating Multi-Service SDNs is now available on demand. Heavy Reading Senior Analyst Sandra O’Boyle and our CTO Cengiz Alaettinoglu gave a great presentation about the promise of SDN in the WAN as well as the challenges facing service providers who want to deploy it. Here’s a quick summary.
Sandra began by discussing the business drivers of SDN as informed by Heavy Reading’s research and discussions with operators. According to their SDN/NFV Tracker for 2016, SDN is a high priority for service providers. The number one driver is greater agility in terms of adapting to service demands. Related to this is the need to deliver new services faster and accelerate innovation to stay competitive. Reducing costs is a factor, but not the biggest one.
One of the reasons service providers need a more automated and intelligent network operations model is the complexity of running multi-service networks. Because applications and services have different performance requirements, fault-tolerance characteristics, and growth rates, running multiple ones on a converged network is especially difficult.
SDN can help address these challenges, but there are many management challenges. As Cengiz explained, SDN controllers enable application programmability, but operators are no longer in charge of the changes. What then governs whether or not these changes should be made? Operators lose visibility into the network. Automation is great when everything works, but when things go wrong, where do you start the diagnosis?
In addition, today when service providers provision a new service, application or a new customer, a planning committee gets involved to ensure adequate bandwidth, class of service, etc. This takes time and prevents service providers from being very agile, but they know everything will run well and meet SLAs. However, if software is doing this, how do you plan for it?
As Heavy Reading has found, it is paramount to avoid fragmentation of visibility and control, especially in a multi-controller environment. What’s needed is to empower software with the same networking know-how and analytics that engineers have to make these decisions.
According to Sandra, this requires “a common management intelligence platform that’s independent and can drive SDN programmability and automation around the areas of policy assurance, performance and analytics.” The platform would provide “…a real-time model of the network that can dynamically calculate the changes needed to either create new services or to maintain the SLAs of existing services.”
This is what Packet Design has done with our SDN Platform (see diagram below). We’ve spent more than a decade developing a rich set of analytics such as routing control plane data, performance data, and traffic data. Applying this to SDN management enables our customers to do things such as balance BGP peering traffic, set up traffic engineered paths, and maintain visibility into the network (including overlay/underlay correlation). It’s an intelligent orchestration platform that can make changes in real time to the network without human intervention, realizing the vision of a self-healing, self-optimizing network.
Register for the on-demand version of the webinar to hear more about SDN analytics, management and orchestration, and a use case: http://www.packetdesign.com/resources/webinars/policy-aware-man…lti-service-sdns/