Network functions virtualization (NFV) technology started with service providers trying to achieve IT simplicity, agility, and cost reduction by decoupling network functions such as routing, firewalling, and load balancing from dedicated purpose-built proprietary hardware. With the performance improvements that have been achieved on standard Intel x86 architecture servers, there has been significant interest in NFV as a viable technology for the enterprise data center. NFV gives enterprises a highly flexible and elastic service delivery mechanism for modern applications that rely on short development cycles, API-driven automation, and elasticity. It frees enterprises from having to buy a physical appliance for each network function and enables them to share resources efficiently, build scale-out architectures, reduce space and power requirements, and simplify operations.

NFV requires IT teams to have the expertise to understand virtualized network functions (VNFs). The challenges of patch management, server maintenance, and configurations across many virtualized servers are daunting to even the most skilled IT teams. New data center NFV strategies are now emerging and are designed to simplify the delivery of this technology for mainstream enterprises so they can lower costs, deliver agile services, and automate operations.

One of most common network services that is ripe for innovation is L4-L7 services offered by load balancers. Load balancing services have long been delivered with customized hardware appliances due to concerns of throughput and performance. However, with load balancing strategies that match the software-defined architectures of data center NFV technologies, there are significant improvements in the way that network services are being delivered. These systems have the five key attributes.

Photo Source: Cisco

Enterprises can now deploy high performance network services on standard Intel servers. The processing power of Intel architecture servers has continued to improve in accordance with Moore’s law while costs have fallen. With special optimizations in packet processing from the Intel Data Plane Development Kit (DPDK), high speed network interface cards (NICs), and processor and memory improvements, high performance load balancing services have become possible at a much lower price point. These technology improvements have eliminated the need for specialized crypto/secure sockets layer (SSL) offload ASICs. The data center NFV platforms can be assembled with Intel servers, kernel-based virtual machine (KVM) running Linux, and high speed 10G or 40G NICs. In addition, software-defined architectures with separate data and control planes for load balancing enable the deployment of highly available network services with multi-tenancy and application isolation. Many of these webscale capabilities that were mostly used by cloud giants such as Google and Facebook are now reaching mainstream enterprises.

More enterprises are adopting agile DevOps practices to develop and roll out applications. Application teams are developing cloud native applications with microservice architectures and containers. The resulting decomposition of monolithic applications causes a proliferation of endpoints in the data center and the associated increase in east-west traffic between applications. Software-defined load balancing combined with data center NFV is particularly well suited for such applications since the load balancing data plane can be deployed on a per app basis if necessary. The simplicity of acquisition, lower cost, ease of maintenance, and configuration for these services enable application teams and operations teams to collaborate on service delivery. The platforms can also make all services available through representational state transfer (REST) APIs to facilitate automation of common tasks and drive self-service for application developers.

The software-defined architecture of these platforms enable administrators to control the system centrally. This presents significant operational improvement compared to managing disparate single-purpose appliances for each network function. Central monitoring and management means that patch management and configuration updates are simplified and can even be automated through the use of REST APIs and orchestration tools.

With a software-defined architecture and central management, IT departments can now gain full visibility of the data plane. The load balancers deployed in the data plane on top of the data center NFV platform can function as a distributed services fabric collecting and transmitting application telemetry to the central control layer. A central controller, which represents the control plane, processes the telemetry in real-time to generate application analytics that administrators and even application developers can use to gain insights into application performance and end-user experience. For example, the load balancers can analyze the traffic to report network latency, security details, including SSL versions and certificates used, and end user parameters such as browsers used, location, and device type. The architecture is ideal for leveraging the strategic location of the distributed load balancers for meaningful insights.

Since the data center NFV architecture provides a modular approach to resource allocation, enterprises can finally gain the benefit of “Lego-block” scalability and elasticity by simply putting additional x86 servers into service. This represents a big change from proactively over provisioning custom hardware appliances for firewalls, load balancing, and other network functions. In addition, due to the API and analytics-driven architecture, decisions on scaling network services can be fully automated using real-time data and orchestration and scripting tools for provisioning resources dynamically. As an example, new load balancers can be automatically created and deployed on-the-fly, based on traffic thresholds and scaled down when traffic recedes.

Data center NFV platforms combined with software-defined load balancing are finally enabling network and IO teams to realize the full potential of automated network operations with more application and DevOps-friendly technologies. This is empowering enterprises to save Capex costs and deliver a flexible framework to deploy virtualized network functions including business-critical load balancing services that address both traditional as well as new east-west traffic management scenarios.