22 November 2018
Next year will be the 50th anniversary of the first moon landings, and the historic ‘giant leap for mankind’, which expanded our earthly horizons forever.
Sharing the same birthday landmark is also the very first Wide Area Network; conceived to a backdrop of Jimi Hendrix, the Woodstock Festival, and the first flight of Concorde, it certainly emerged and grew up in a time of change.
The first real use of a WAN was by common agreement ARPANET, which connected together four US academic campuses in 1969, using packet switched technology which was revolutionary in its time, and many of the fundamentals set down are still relevant today.
WAN technologies have always been evolving, from leased line to Frame Relay, to Asynchronous Transfer Mode, to today’s predominantly Multi-Protocol Label Switching networks, via x25 and other protocols on the way.
In the WAN world long lead times on circuits, typically 90 working days, coupled with inflexible ability to change network topologies, and dynamically adapt to a changing business model in the real world, feel very much like a 20th Century throwback.
But in the last few years technologists have been driven to think ‘cloud first’ and ‘as a service’, as businesses strive to consume opex services, rather than tie up capital and pay bank loans to finance set up costs; a new generation of IT managers don’t see outright ownership as necessary, and want to buy in IT and Telecoms in the same PAYG way.
Like all new disruptive technologies it needs assessing dispassionately against a background of over expectation, and with a bit of prudent caution, but SD-WAN can be thought of now as ‘WAN2.0’, being a breakthrough in its own right rather than a simple progression of the past.
SD-WAN means that in a radical way networks can now de-couple the WAN control plane technology (the ‘thinking’ layer, with management UI and configuration data, routing policies and decisioning) from the physical data plane (the ‘doing’ layer, which moves the data of the applications around), and deliver highly flexible networks, often mixing traditional MPLS and lower cost internet access broadband circuits. A single logical instance of the control plane can manage multiple instances of the data plane.
The modern WAN needs to facilitate bursty and on-demand traffic, and have carrier grade performance for inter-datacentre and core sites, plus be flexible enough to allow edge and branch sites to be accommodated rapidly, and set up and torn down at will. Many firms will use cloud compute in lieu of traditional datacentres, so that fits SD-WAN very well too.
Cloud services such as Salesforce, Office 365, Skype for Business, and hosted SaaS offerings generally, are prime drivers for SD-WAN as it can remove the concept of all cloud traffic routing in and out through a corporate data centre with clever use of route optimisation.
Today’s networks are changing dynamically every single moment, with new virtual machines being spun up, the demand for rapid access for BYOD and mobile, and the use of multiple micro-services and throwaway apps, in a way that is very unpredictable.
Using old fashioned and high cost fixed line circuits is no longer viable, and their displacement by low cost direct internet access or DIA circuits, and broadband services, in a mix and match way with MPLS and other protocols is the new paradigm.
This means that heterogeneous networks are very viable, with a mixture of MPLS, DSL, LTE and cable if so desired, rather than traditional homogeneous networks. There are also real and significant reductions in management and administration costs associated with de-coupling of the management layer, so that there is no need for per device change and administration. SD-WAN can leverage and supplement existing legacy technologies, so it isn’t a full rip and replace from the ground up either.
The other great benefit of a centralised management layer is the core ability to control network performance and characteristics to compensate for packet loss and jitter, manage congestion and alternate route selection. The management layer uses SDN Controllers as the policy definition hosts, and these prioritize traffic routing requests for optimal efficiencies, often being coupled with WAN acceleration for an added boost.
This is especially relevant for real time and synchronous traffic (video, voice and collaborative technologies), and can be managed with granular application specific policies, in conjunction with path selection, tunnel bonding, use of virtual WAN overlays, security hardening using TLS encryption, inbuilt cloud intelligence and Internet breakout, and path conditioning functions.
SD-WAN can make clever use of asymmetric links that appear congested with poor RTD figures, so it’s a ‘more bangs per buck’ technology too.
As businesses have an increased need to access rapid access caching layers, big data lakes and large scale cloud elastic compute services such as AWS, there is a concomitant need for improved management of QoS too; get this wrong and a degraded user experience will result in customer dissatisfaction, poor mean opinion scores, and ultimately lost revenue.
Adaptive, resilient and self-annealing SD-WAN networks are the future, and will displace the hegemony of MPLS in coming years. SD-WAN is on track to be a $1.3Bn business by 2020, and is of huge significance to the telecoms industry globally.
It’s entirely possible that in another 50 years, and on the 100th anniversary of the Apollo 11 landings, with Neil Armstrong’s words still echoing down the years, we will have Inter-Planetetary Networks or IPAN, but for the here and now, SD-WAN is the future.
Gary Dudbridge is a guest blogger for Gamma, with extensive experience as a Lead Telecommunications Architect with telecommunications networks, contact-centres and systems within high profile multi-national organisations, especially in the financial and telecoms sectors.
Gary Dudbridge Social: Linkedin