2016 was a big year for Software-Defined Wide Area Networks (SD-WANs). I expect that it will continue its market growth for at least a couple of years more. SD-WAN is built on Software-Defined Networking (SDN) technology focusing on secure WAN connectivity between branch offices. It redefines the network from autonomous packet-based routing to application-based routing and business-grade security to enable Internet circuits for enterprise WANs.
The main business drivers for SD-WAN that appeal to enterprise customers are:
- Lower network costs to replace MPLS to Internet
- Better performance, resiliency and visibility using dual active links
- Ease of deployment and management by a centralized controller
However, whilst you will realize benefits, in practice and with experience, you will find that it will take time for SD-WAN to mature into the elegant solution that realizes the levels of cost savings (30% or more in some cases) that many of the vendors are touting in their marketing.
So, you have decided to investigate and potentially implement SD-WAN, but how should you approach this? What might the future hold? Below I outline some of the issues that you will face and I suggest some areas for improvement (by vendors). And then I gaze into my crystal ball.
#1 Develop interoperability to provide flexibility
Carefully plan your selection process: research the SD-WAN vendors, produce a short list, implement proof-of-concepts with each and then select your (single) SD-WAN vendor of choice. There is no compatibility between SD-WAN vendors because their focus is on WAN connectivity and the elimination of the complexity associated with the management, configuration and orchestration of WANs. You will distribute hardware to all locations, if you are lucky you may have the option to host SD-WAN software on your servers, but most likely you won’t have this choice in all branches. That’s tough if there are hundreds or thousands of sites in your estate.
Next you will think who will deploy and manage it. You may have the skills and capabilities in-house or you might contract with a service provider or integrator. It might be you are asking managed SD-WAN as-a-service vs 3rd party overlay to discuss the available options. Be warned that once deployed, it’s not easy to swap to an alternate SD-WAN vendor — you will likely need to rebuild from scratch. Please remember that it is not a traditional network with the associated levels of maturity across the eco-system. Build the time into your process for careful selection of your chosen vendor and where appropriate, service provider/integrator, check the relationship between these key partners, consider vendor roadmaps, take references and be sure that they have delivered real market implementations.
As the technology develops I will expect that standards will develop and this lock-in will be removed providing the customer with greater eco-system flexibility.
#2 Host Virtualized Network Functions (VNF) instances and integrate with SD-WAN to complete the WAN solution
Today’s WAN network is not simple. You will have cloud applications, firewalls, IPS, IDS and other security appliances, remote access and more. SD-WAN does not fully cover all network functionality. During the planning and design phase you will need to give careful consideration as to how you can integrate SD-WAN with your current WAN as well as your wider infrastructure.
Along with SD-WAN, Virtual Customer Premise Equipment (vCPE) and White Boxes are another network trend for branch connectivity. VNF instances run in a single physical device, which could include SD-WAN virtual instance, however, it is not perfectly integrated into one single controller; you would need to install, setup and manage each VNF instance separately; managing these through a single controller and portal is not provided by vendors as of now. This is a functionality that I would expect vendors to have in their development roadmaps – which reverts to #1 above – be sure to include this (future roadmap) in your vendor selection process.
#3 Cloud connectivity to support a cloud-first strategy
Analysts, the media and experience all point to an accelerated adoption of a cloud-first business strategy by enterprises. There is an increase in: the number of servers migrated to the cloud, and legacy application licenses that are being moved to cloud-based SaaS models. As such cloud connectivity is a critical business consideration and feature. Some SD-WAN vendors have been developing native support for secure cloud connectivity and this will expand to encompass cloud and SaaS providers along with SaaS optimization.
The problem I often see is that the customer doesn’t initially understand the criticality of the network when developing this cloud-first approach - it is almost an afterthought - then they are surprised and need to catch-up. To implement SD-WAN correctly you need to fully understand your environment and how your business uses the infrastructure, such as; business applications and associated priorities, traffic patterns and flows, what cloud services are in use, SaaS models, and how and where to connect to these.
Uncertainty can often arise in agile application development environments and where there are agile dynamics in a business, it is not necessarily easy to understand all of the services that are connected to a modern network. Failure to complete this analysis, to understand or identify how the business uses the network, can result in additional requirements post SD-WAN deployment. This can then require a rework of the entire design and result in a messy, patched network. Again go back to #1 above, include this within your selection process and identify a partner that has a track record and can evidence this level of understanding and analysis.
If you do not have the right level of detail, you need to create the time to source it. Time spent on this research and auditing during the planning and design phase will prevent many a sleepless night in the future. An SD-WAN implementation integrates with the fabric of the business; it does not simply connect points. SD-WAN planning and design requires more time and conversation with internal (and external) customer sets, a deeper understanding of the business, when compared to traditional WAN network planning and design.
I would expect SD-WAN vendors to more natively support and provide simple and secure connection to any cloud and SaaS services so that you don’t need to redesign the network every time a new requirement comes along.
#4 Big data analytics to analyze your network
SD-WAN provides more granular configuration based on each business application. It’s a great feature to have only if you already know about all business applications. In most cases it’s difficult to get such information from customers as they often don’t know it fully. Also “cloud-first” approaches add more and more applications on the WAN and require tuning the SD-WAN manually.
I expect that big data analytic is the solution to solve these issues. Real-time traffic data, access logs, system and error logs, usage, latency and much more for a variety of network devices will become crown jewels. Big data analytics could, in the not too distant future, be applied to support SD-WAN solution design and might aid: efficient network path control, capacity forecasting and automatic scale-up, performance reporting, security protection and so on. Also, it will improve operational efficiency and overhead. Current network operations are still based on detect and response. SD-WAN could potentially increase complexity for the operations team because it will create an additional application layer. A big data solution could support predictive analytics and it will be able to predict issues and enable response and resolution prior to disruption.
#5 Artificial Intelligence (Machine Learning) infused into SD-WAN to automate networking.
One of the advantages that SD-WAN provides, is easy or zero-touch delivery by a centralized controller in the cloud. But it doesn’t mean you don’t need to understand your network. It doesn’t mean you don’t need to touch the network once deployed. It just simplifies configuration(s). To set it up correctly and get it to work, you need to know your network (and how business wants to use it) in detail. You still need to change configuration manually, based on requirements. As cloud adoption progresses, changes will happen even more often. I’m hoping #4 and Artificial Intelligence (Machine Learning) could automate and eliminate these bothersome requests.
Artificial Intelligence (Machine Learning) is also a hot topic these days – generating a lot of coverage in the media such as driverless/autopilot cars – more start-ups attracting investment. I can foresee this evolving into the field of networking: Big Data + AI = Network Autopilot, or Cognitive network another round of start-ups and another round of investments! In all seriousness it is and will become possible for a machine to analyze all network data and applications and for a machine to decide the best practical network setup for the customer — and to constantly monitor, real-time configure/adjust and optimize performance — auto-predict issues, bypass them, raise tickets. Hmm, I think that I am going to found a start-up?
In summary, Enterprise cloud-first strategies are driving an ever growing need for cloud services: computing, storage, data backup, machine learning, big data analytics and more. This is causing enterprise to reconsider the network and how it integrates with cloud services and the fabric of business. SD-WAN is missing pieces to provide best performance to cloud connectivity and to bring customers an end-to-end programmable infrastructure. There are still lots of pitfalls and problems; vendor lock-in, compatibility, orchestration, automation and more. Things needs to improve a lot in the SD-WAN space to turn it into an elegant solution. I think it will be happening soon as Cloud-services and SD-WAN are strongly bonded and interdependent- I expect a continuous sustainable innovation to fix these issues soon.