5 minute read

Future of Network Infrastructure - on Cloud

By Shantanu Bhattacharya

History and market pressures

Some time ago, the primitive concept of Software Defined Networks (SDN) was brought to the fore. Proponents of SDN claimed that network device vendors were not equipped to handle the changing speed demanded by the industry. Programming the devices was only possible through the CLI or using the Simple Network Protocol; neither met the evolving requirements for easily accessible, flexible, and application-friendly interfaces. That led a few Stanford University engineers to create OpenFlow protocol, enabling an architecture comprised of a number of devices containing only data planes to respond to commands sent to them from a logically centralized controller that held the control plane. The controller was responsible for keeping track of all the network paths, as well as configuring all the network devices it controlled. These communications were the essence of the OpenFlow protocol.

OpenFlow helped in conceptualising the SDN. OpenFlow could “transmogrify” these platforms to be any network device. E.g., firewalls or NAT. This dramatic shift in the networking industry was well documented. Network Functions Virtualization (NFV), a related concept, has allowed commodity servers to accomplish the roles of the network devices. That significantly reduced cost and speed of service deployment. In an NFV, the virtualization layer operating system coordinates the compute and store. Further, it connects resources shared among the Virtual Network Functions (VNFs) that could then execute on the same physical server. The Management and Orchestration (MANO) component orchestrates and administers the VNFs. With ever-increasing demand for network bandwidth and services, virtual functions can be deployed on demand.

NFV and its evolution

Network Function Virtualization (NFV) could be used for some basic and prime concepts of SDN. That included control/data plane separation, logical centralization, controllers, network virtualization (logical overlays), application awareness, application intent control, and many more on easily available (Commercial Off-The-Shelf (COTS)) hardware platforms. NFV has enhanced the conceptualisation of new methods in support of service element interconnectivity, and techniques that can cope with its dynamic requirements and their upscaling and downscaling.

The market pressure on network operators increased in 2013 and real challenges were posing their businesses. • What started as an Over-The-Top (OTT) video and social media into their broadband customer base, grew into OTT service offerings. The outsourcing of organisations’ IT to cloud providers turned these new competitors as more relevant IT partners. • Wireline operators, faced large and long-delayed transitions in copper-based services. • On another front, virtualization concepts evolved out of enterprise-centric virtual machine operations, to more composable and scalable components like containers

in public and private clouds. • Virtualization infrastructure performance optimisation started receiving a massive adoption through efforts like Intel’s Dataplane Development Kit (DPDK)enabled version of Open vSwitch (OVS)— resulting in throughput increase of virtualized network functions more easily achievable. • Cloud computing attracted more and more enterprise customers. In addition to COTS costs reduction, this also created an environment for more service outsourcing.

Virtualization does not work for all service deployment problems and actually introduces new reliability issues that a service orchestration needs to mitigate. Virtualization, like any tool, deployed wantonly, can lead to disastrous consequences. While virtualization is critical for NFV, the orchestration and integration required, needs a scope that includes present and future fully integrated service platforms. In that regard, SDN can provide a “glue” for enabling middleware.

Even though SDN is universally accepted in the control of service virtualization, the type of control point or points are still debated. For NFV, the debate manifests around stateless and proxy control points, or when inline or imputed metadata is employed.

Good or bad fit?

NFV merits

There are many benefits of implementing NFV as an alternative to standard architecture. Some of them are discussed below: • Reduced CAPEX and OPEX • Flexibility of scaling the network up and down • Enhanced service agility to support faster service rollouts • Increased operational simplicity • Speedier innovation leading to eliminating hardware change • NFV can be a viable revenue generator. • “While challenges in using NFV to impact service velocity remain, the technology can enable a new range of service features that can produce revenue,”

Tom Nolle of CIMI Corporation opined. • Increased data collection, analysis and business decision-making processes.

NFV demerits

NFV is based on SDN and thus has the same limitations. Like SDN, NFV needs to evolve and become reliable in enterprise level deployments. However, it is close enough. Some specific challenges facing NFV going forward include: • Manage both cloud-integrated hybrid environment and physical devices • Differs from conventional IT environments in that NFV requires abstract IT management • NFV needs to be more dynamic than traditional environment for it to be useful • Process realignment for the simultaneous management of traditional and virtual infrastructure is required.

Approach for mitigating challenges

Wide adoption of NFV will eventually happen. It would be adopted for some use cases due to it being the need of the hour. Legacy networks will remain for some more time.

Meeting the requirements of a transition towards virtualization, requires the architecture to allow: • Supporting dynamic, real-time network and service changes required by network events • Separating network configuration and management of network state • Supporting a modelling approach to network services • Interworking with network orchestration platforms and

SDN controllers

It’s important thoroughly plan the migration strategy before one begins deploying network virtualization. It might be more relevant for large enterprise networks. Replacing existing infrastructure can be much more complex in these cases.

Adopting hybrid environment where the virtual networking capabilities are deployed in the areas where they offer the most perceived value while allowing legacy in others might be the way forward.

NFV is new and is in vogue. It provides the promise to handle many of the challenges of current and future networks. But, like all innovations, scrutiny is suggested along with incremental take up to verify provided value before deploying widely.

About the Author Shantanu Bhattacharya is a software professional with around 25 years of diversified experience.

Proactive leader, known for steering multi-million dollar endeavours to deliver cost savings and extensive profits by impressing stakeholders and on-boarding them. A professional with global experience in large multi-national enterprises, specialising in harmonising human capital by stimulating innovation. Has brought about organisational change and large transformations in trying circumstances to boost profitability.

Have extensively designed and created architecture for application software for retail, system integration, and healthcare; networking software that's SNMP-based, and TCP/IP stack; security software; a file system for India's first supercomputer; and Real Time Software for the Indian Missile Program. Has published papers (external and internal) in various forums and has been a reviewer for ACM Computing Surveys.

While virtualization is critical for NFV, the orchestration and integration required, needs a scope that includes present and future fully integrated service platforms. In that regard, SDN can provide a “glue” for enabling middleware.