Compute infrastructure technology has undergone a fascinating evolution over the last decade. Virtualization changed the way compute is utilized and the way IT operations are run. Cloud is now changing the economic model of IT infrastructure and bringing compute closer to application developers and business owners. Meanwhile, connecting everything together and enabling the dynamic nature of modern compute, the network looks due for a revolution of its own.
Software defined networking (SDN) promises to unleash the power of the network and turn it into an efficient highway system, allowing applications to run optimally without requiring organizations to overinvest capital and operations in their network infrastructure. SDN is shaping up to be to the answer to the increasing demands of both the dynamic data center and mobile enterprise users running data-hungry applications.
Networking: Past and Future
No longer standing back and waiting for the verdict on whether SDN will become mainstream reality, many network equipment manufacturers are now communicating SDN roadmaps. Some have even released SDN products that allow their existing network devices to be managed by an SDN controller, to an extent relieving them from their former duties to think independently and automatically react to changes. SDN simplifies the life of the network device and requires it to only obey the network controller. This will likely reduce the prices of individual network devices.
Meanwhile, networks themselves were not designed and built to be software-defined. Standards bodies like the IEEE and the IETF have worked hard to define various interoperability standards so that networks can run as much as possible by themselves. Among these technologies are link discovery protocols, topology management protocols and routing protocols. Within the domain of a single SDN controller, these protocols become obsolete. Their main benefit in future will lie simply in connecting SDNs together.
Theoretically, buying inexpensive network devices for management by fancy network controllers will get you a fancy network. Whether that will be the case is still up for debate, though. There are not enough commercial SDN deployments yet to be sure. It is especially hard to tell when it comes to the operations required to run a SDN. But despite these reservations, there are other compelling arguments in favor of the adoption of SDN and its close relative, NFV.
The Technical and Business Arguments for SDN and NFV
Enterprises need a compelling reason—not just a vendor pitch–to commit to implementing SDN. Most likely, that compelling reason will come in the form of a problem the existing network cannot solve, or that it cannot solve within a given budget. One of the potential benefits of making the SDN programmable and capable of performing advanced network functionality everywhere is the ability to implement application network services, such as security, acceleration, and load balancing, throughout the network. Dynamic workload mobility in clouds and bandwidth management over WAN links are probably the most SDN common use cases. Protecting applications everywhere in the network and extending other network services to the entire SDN are also possibilities.
Business agility is another compelling argument in favor of SDN. Currently, enterprise customers have to make hard choices once they deploy application delivery and security services in their data center networks. With legacy networks, the way they choose to implement these services initially is usually the way they will need to continue implementing them in the long run. Today's static network attachment model doesn’t fit today's dynamic business environment. The ability to dynamically program the network to handle ongoing changes to application infrastructures, while continuously delivering network services without the need to redesign the network, makes more sense. Applications and network devices that communicate with an SDN controller will drastically ease service delivery throughout their network.
SDN allows and supports network services in ways that legacy networks cannot. Various operations that previously required human intervention can now run automatically. Some of these operations include: increasing the capacity of existing services, deploying new services, ordering services in a certain sequence, and maintaining services. Along with SDN, Network Functions Virtualization (NFV) addresses these challenges.
SDN vs. NFV
Is NFV independent of SDN? NFV, like cloud computing in many ways, encompasses a set of technologies that have matured enough to potentially enable new solutions not possible today. NFV is the collective use of the following three technologies:
Server virtualization to host network service virtual appliances as efficiently as possible to enable high performance of network services.
SDN to program network traffic forwarding, seamlessly delivering network services in path with required availability and scalability attributes.
Cloud management technologies to provision network service virtual appliances and orchestrate connectivity to these services by managing the SDN, enabling functionality of these services by managing the service itself.
Network services typically significantly impact applications' quality of experience. Automating these services with SDN improves QoE due to the programmatic, predictable nature of SDN. Layering NFV on top of the automation of an SDN network and utilizing SDN applications that program the network to offload basic network service functionality will improve network utilization by automating traffic path selection and network service provisioning and de-provisioning. Due to the high network utilization, the value of investment in SDN, SDN applications, NFV frameworks and SDN compatible network services can be greater than equivalent investments in traditional networks and network service appliances. Today’s customers expect solutions in both categories.
SDN and NFV: The Economic Argument
Carrier customers often have to make hefty investments in network service appliances based on projected future service utilization estimates. Often, investments end up unjustified or need to increase as service adoption dictates. Neither case is favorable. A heavy investment thrown away is wasteful, and new investments to increase service capacity require a new network design involving significant operational overhead for planning and implementing the capacity increase.
Customers demand solutions that can easily be integrated into the network and scaled in the form of NFV. A network service management framework that constructs an NFV platform consisting of an NFV application and a network service virtual appliance, for application delivery and security that allows them to automatically scale services in their networks, is a promising way forward.