There is much confusion about the differences between Software-Defined Networking (SDN) and Network Functions Virtualization (NFV). Both network traffic management strategies provide some form of virtualization for network services, abstracting software controls from hardware resources to increase network performance and versatility. SDN and NFV overlap in many areas and are complementary technologies, as well as competing approaches to solve the same problems. In the context of big data, SDN and NFV can work together to optimize data access and improve overall analytics performance.
Infonetics Research predicts the combined market for SDN and NFV products reaching $11 billion by 2018, with NFV products commanding the lion’s share of the market. The analyst firm sees the combined SDN/NFV market as three distinct categories:
- New SDN and NFV software, which makes up 20 percent of the market
- Displaced revenue from companies buying NFV hardware and SDN solutions instead of something else, which is about 12 percent of the market
- Revenue from new segments of existing markets looking to buy virtualized network functions and SDN-capable hardware, which will be 68 percent of the market in 2018
For big data applications, both SDN and NFV technology play a part. The key is to understand the applications and limitations of each platform.
Distinguishing Between NFV and SDN
SDN and NFV provide similar functions but are different in their approaches. The differences are important.
SDN is a broad term that describes ways to make a network more flexible and agile by uncoupling the network control layer from the hardware infrastructure. As a result, network engineers can manage a multi-vendor network fabric without needing to deal with specific hardware management systems. The advantage of SDN is that network configurations can be changed quickly and easily. Components such as data storage can be added as needed, and network traffic can be changed from a central control console—both important capabilities for big data performance. SDN also enables network virtualization and programmability of network services.
NFV also is designed to virtualize network functions that would normally be handled by proprietary network hardware, such as routers, firewalls, and load balancers. Network services are decoupled from the networking hardware and hosted on virtual machines (VMs) under the control of a hypervisor. The advantage to network administrators is network traffic is managed in software so switches, servers, storage, and even cloud resources can be managed using VMs rather than more expensive, dedicated hardware.
Understanding the evolution of SDN and NFV is also useful. SDN is a concept developed in the data center to make network devices programmable. By separating the control and forwarding functions and centralizing network control, enterprise behavior can be programmed and even automated. As a broad-based initiative, it was clear that SDN standards were needed, so the Open Networking Forum devised OpenFlow, a model to define how traffic is organized. The OpenDaylight Project also is working to develop open standards for SDN.
Network service providers devised NFV to speed up deployment and optimize new network services, which increases their revenues. Proprietary hardware appliances were impeding adoption, so NFV addressed the problem by virtualizing network functions. The European Telecommunications Standards Institute (ETSI) oversees the basic requirements and architecture of NFV.
Complementary, Not Competing
The core objectives of SDN and NFV are to improve network performance, and they use similar approaches:
Both move functionality from hardware to software.
Both rely on commodity hardware rather than proprietary systems.
Both use simple APIs to deliver visibility into network functions.
Both support more efficient virtualization and automation of network services.
Despite these similarities, SDN and NFV are not competing strategies but, rather, are complementary. You can deploy one or the other, but when you use them together, you improve network performance even more. SDN uses policy-based decisions to manage network traffic and promote network automation, while NFV focuses on network services aligning capabilities with the virtual networking environment. The result is better overall network performance and more efficient traffic management.
Addressing the Demands of Big Data
Optimizing network performance benefits big data. To be effective, big data applications must be able to access data quickly for analytics, so network speed matters. However, while both NFV and SDN support virtual networking and improve performance, SDN has some additional capabilities that offer real benefits for big data applications:
Bandwidth management – Combing through large datasets using parallel processing means more data bandwidth. SDN helps by managing connectivity and optimizing throughput.
Dynamic cloud computing – The cloud delivers capacity and resources on demand. SDN is better suited to dynamically access cloud services and storage as needed from within the data center.
Managing data traffic – SDN has dynamic routing tables that prioritize data traffic, which simplifies management of multiple virtual machines and IP addresses.
Scalable and agile – SDN also makes it easy to add new devices as needed. In the case of big data, this means meeting the demand for more data storage and computing power from the cloud.
Policy and security management – SDN also supports policy management across the entire infrastructure, not just for specific devices.
The question isn’t really whether SDN or NFV is better for big data applications, but, rather, how each can contribute to improving network performance to support big data analytics. SDN certainly has value for big data applications, but adding NVF to the network mix won’t hurt big data performance.