In a survey published by IDG and sponsored by Masergy there is a clear indication of growing concern over the cost of SD-WAN deployments. The survey showed that cost is the number one concern of enterprises and that this concern has slightly grown between July 2017 to May 2019.

This concern may leave some puzzled since many view SD-WAN as means for cost reduction compared to MPLS, how come that the cost of SD-WAN is now perceived as an issue or something to be concerned about?

Welcome to reality

First and foremost, it is important to remember that the main driver for SD-WAN adoption is cloud and digital transformation, moreover, many enterprises that never used MPLS or used it modestly, are moving their networks to SD-WAN.

This shift, combined with the move of services to the cloud and the natural growth of data traffic means that the total bandwidth consumed by enterprises is on the rise. Given the cost structure and business model of most SD-WAN vendors and providers, cost becomes an issue.

For example, a mid-size enterprise that deployed SD-WAN of a leading and well-known SD-WAN vendor saw bandwidth grow by a factor of 5 since they deployed the system in their network. An IT leader of that company told me that the increase was due to 2 factors:

• SD-WAN itself – the combination of duplication of traffic and SD-WAN overhead caused an X3 increase of bandwidth consumption
• Natural growth of traffic due to new applications and consumption of cloud applications

It’s their business model stupid

SD-WAN has nothing to do with open networks. Yes, SD-WAN solutions/vendors are heavy users of open source, this doesn’t imply to their solution being open. On the contrary, most SD-WAN solutions follow the footsteps of traditional networking vendors – as closed as possible and complete as possible, by that, the vendors create 2 layers of lock-in for their service provider and enterprise customers:

High switching cost (not packet switching but rather vendor switching) – The model can be compared to the HW+OS model of compute devices. Compare what aunt Marry used to do with her PC 5 or 10 years ago. Pretty much the same tasks as she does today, read/write documents, email and web. But good luck with trying to run a current OS on a 5 or 10 years old PC. In networking products, we see a multi-year commitment lock-in model that combines HW+SW with a constantly growing SW stack that increases HW requirements. The vendor in-turn offers options for HW upgrades that come bundled with the extension of the lock-in period.

Keep other vendors out – Providing a big SW stack that spans many disciplines and capabilities that have no technical reason to come bundled in one package, lock the door to smaller vendors who offer only mission-specific applications. In SD-WAN we see vendors include DPI, Security, WAN optimization, network monitoring and other capabilities in their products. Although most of these capabilities are offered also by vendors that focus on that function only and are experts in that field. There is no option for the SP or enterprise to decide which vendor to use for a specific capability, it’s a “take it or leave it” deal.

In this sense, the more SD-WAN expands in the capabilities it offers becoming the swiss army knife of networking, the stronger the vendor lock-in is.

SW stacking increases cost

Clearly having a one-stop-shop has it pros, especially for the smaller enterprises. On the other hand, we see a large percentage of those smaller enterprises that don’t really want or need all of these capabilities.

SW stacking comes with a toll. The larger the SW stack is the higher the licensing cost will be, additionally, larger SW stacks impact HW requirements and performance. The result is a higher Total Cost of Ownership (TCO).

Modular SD-WAN

The other and more cost-effective option is a modular SD-WAN that allows not only to add and remove capabilities but also to select from which vendor to source some of the capabilities.

This architecture is possible by creating a 2 layers architecture in which the base layer is the networking infrastructure while the second layer is the add-on capabilities that integrate as application.

Networking infrastructure

This layer is the foundation of the network, it allows to connect between the different branches and from the branches to the cloud. It allows to open secure connections (tunnels) between the different sites and manage them from one central management system.

This layer would also include the infrastructure for installing and running those applications that provided the added value capabilities.

Application layer

This is not a VNF based integration but rather an integral part of the SD-WAN solution. This means that the applications run in the router’s data flow or integrated into the management system. If we take as an example an application that needs to manipulate packets that runs as a VNF, the flow will be:

• Data enters the router and processed
• Based on the forwarding graph it needs to be handled by the VNF application
• Data exits the router and is sent to the VNF
• After the data is processed by the VNF it returns to the router
• Then the data is encrypted and sent to the network

On the other hand, when the application runs in the router and is part of the data flow, it is simply another node in the forwarding graph.

Most SD-WAN solutions are built in a closed architecture that doesn’t allow for 3^rd party applications to run inside them. Moreover, in most cases, even the option of the VNF is limited because traffic exits the router already encrypted, hence, there is very little that those service chained VNFs can do with it.

A modular solution allows an enterprise to select only the elements it requires to have in the SD-WAN they deploy while creating a more competitive environment as there are multiple options for different 3^rd party applications that run in the solution.

Service providers can have better control and differentiated services so for example, a service provider could offer different capabilities and features depending on the segment to which it offers it’s SD-WAN service to. A service provider could offer only the baseline networking infrastructure to very small enterprises while adding more features to larger and global ones.

By this, it will be able to have a flexible cost structure that also impacts the competitiveness of the pricing it would offer to each segment. Moreover, it would also be able to integrate its own applications that improve the quality of service of its applications and create a tighter relationship between the SD-WAN overlay and the underlay of the network.

In July, flexiWAN released its first beta version. flexiWAN (which I’m CEO and co-founder of) is on a mission to democratize SD-WAN. We are doing this based on 2 major differentiators that keep us apart from existing SD-WAN solutions:

• flexiWAN is open source. As announced, we will release the code at the end of the year [update – code is available as of January 2020]
• flexiWAN features an open architecture which makes it modular

Interested in kicking the tiers of flexiWAN, open a free account and try it yourself.  

Closing notes

The cost of SD-WAN is still a concern keeping enterprises on the fence with their decision to move forward with its deployment. A major reason for its high cost is the dominance of the market by the large and traditional networking companies that lock-in their customers with large and monolithic SW stacks that put them in the vicious cycle of SD-WAN cost. The way to break out of this vicious cycle is to open the SD-WAN market through open source and modular solutions. This give enterprises and service providers better control and differentiation by selecting the elements and capabilities of the SD-WAN product they deploy which in turn impacts price.

• Learn when standards matter (and when they don’t)
• Learn why appliance consolidation shouldn’t drive your SD-WAN selection process
• Stay away from the SD-WAN cookie-cutter model

The post The Vicious Cycle of SD-WAN Cost appeared first on The New Dial Tone.