I spend a lot of time on the road. Meeting with customers. Talking at industry events. Working with our sales team. Briefing analysts. I love it. One of my observations from all of this interaction is that everyone agrees, for the first time in a while, that things need to change in the networking space. The way we architect our products and our networks just doesn’t match how enterprises and consumers access content and data – and it’s “crimping our style”.
As everyone knows, the datacenter is already an elastic and virtualized resource, which is rapidly migrating towards SDN architectures to further refine the deployment model. The WAN however, is still largely a static environment that is provisioned on a point-to-point basis. The differences between these approaches (in managing and orchestrating the datacenter vs. the WAN) become glaringly evident when trying to deliver end-to-end services that take advantage of elastic cloud resources. Piecemeal orchestration becomes an even larger issue when you consider forward-looking network architectures that include NFV.
Interestingly enough, the cloud has not been a silver bullet for addressing all enterprise IT challenges. In conversations with large enterprises, one of the primary challenges they face with the datacenter-centric deployment model is the lack of coordination and control they have over the end-to-end service or application experience. The quality-of-experience for an end-user, is comprised of all of the individual components that constitute the service, i.e. the infrastructure, the application performance, and the network connectivity. Managing the datacenter resources dynamically and in an automated manner is not sufficient; because the network quickly becomes the constraining factor.
The next-generation of enterprise services require coordinated orchestration of WAN and DC resources, not only OpenFlow-based datacenter fabric provisioning and instantiation of VMs and other datacenter resources, but also provisioning of the WAN connectivity. Furthermore, giving an enterprise customer the ability to fully manage their cloud service from end-to-end, gives them the control to manage the end-user experience, which can often be the differentiator for a business-user that is unwilling to give up the control they have today with in-house IT infrastructure and networking.
A multi-vendor, multi-layer, multi-domain SDN orchestration platform can enable the delivery of traditional cloud services with integrated control and end-to-end awareness. Multi-vendor is about more than physical network equipment, but also encompasses multi-technology, such as cloud-management platforms, deployed applications, and more. Operators are singing loudly-and-clearly that they want to break-free from vertically integrated solutions as part of this transformation.
Piecemeal orchestration of the network and cloud resources, as prevalent today, results in inefficient use of resources, and does little to ensure the end-to-end service is delivered in a coordinated manner and per the required SLAs. For example there may be mission-critical services that requires high-availability or ultra-low latency, while other services — like nightly off-site data replication — may be hosted at a site that does not offer as stringent performance guarantees, but offers lower cost-points.
OK, so let’s take a breath and shift gears. As outlined above, it is my belief that the industry can figure out the disparity between how data center/cloud resources are orchestrated and how the network can complement that – but there’s more to the story. What about NFV? And new open source initiatives?
Let’s start with NFV. The rapid industry movement towards placing network-critical functions in a virtualized environment (see ETSI NFV ISG), makes it even more evident that we have to fundamentally change the way networks are managed and operated.
To realize the promise of network and application virtualization and to enable service-agility, network operators need SDN and NFV orchestration platforms that span both the network and datacenter domains. These tools must support multi-layer coordination as well as allow operator and enterprise-specified policies to influence service, network programming, and resource allocation. These platforms must support intelligent datacenter resource allocation based on application requirements, performance criteria, as well as targeted cost-points. Systems that strive to provide these capabilities, with open-architectures, will allow operators to orchestrate services that span both physical and virtual networking devices and functions (NFV) and ultimately offer the flexibility network operators are asking for.
Another key movement we are seeing in the enterprise and operator community is enthusiastic backing of open source initiatives. Industry initiatives like OpenStack are growing rapidly, and continuing to garner support across the entire ecosystem of vendors, enterprises and network operators. Embracing open-initiatives is a critical component of next-generation SDN solutions. The idea of open APIs, published interfaces and the creation of open communities around these technologies gives operators the freedom to build networks and deploy services using best-in-breed solutions, rather than being locked-in to any given platform, technology or vendor offering.
Open SDN architectures offer incredible potential to transform how network operators build and interact with their networks. With rapid advancements in virtualization technologies, performance gains made by commodity servers, the fusion of IT automation and network management, and rapidly growing open-source software initiatives, we have an incredible opportunity to change how networks and services are designed, operated and viewed.
I enthusiastically look forward to the transformation around us, to the next customer meeting, to the next speaking opp, to breaking all the “old” rules, and to equipping our customers with the tools and platforms they need to build software-defined networks. That’s what makes our industry so great — we don’t change for the sake of change, the network evolves because customers and applications require it.