In some organisations, a strong and immediate case can be made for SD-WAN. IT departments that pride themselves on leading the curve rather than following it will already be actively exploring, trialling or perhaps even in some full deployment. Some see an investment in the underlying bandwidth as a more beneficial approach as is seen with the march to full fibre Ethernet based networks.

Building a use case

The adoption of a cloud-based infrastructure, for both storage and applications, means that demand for bandwidth grows in line with usage of these services, especially if these applications live in the public cloud.

Moreover, key enterprise applications such as Unified Communications (UC) are becoming richer and more sophisticated, placing greater demand on the network in terms of latency and bandwidth.

Video is arguably the biggest driver. The multi device, multi application trend pervasive in business today means that users are consuming more rich media in both a personal and business capacity, sucking up the bandwidth on their smartphones and tablets; placing additional demands on the underlying infrastructure that wasn’t necessarily designed for this new rich mix of competing application types and styles.

But it’s not just consumer applications that are to blame. More and more enterprises have adopted video for conference calling and remote working, and a sudden burst of simultaneous usage can have a knock-on effect on latency-sensitive applications and result in a negative experience all round.

The building blocks:
An SD-WAN solution if properly designed, implemented and managed, should be capable of delivering:

  • Agility for rapid deployments and change
  • Improved network performance for critical applications
  • Cost efficiencies

To achieve these headline gains an SD-WAN solution needs four key features:

1. Hardware replacement

SD-WAN hardware should make dedicated routers redundant by allowing direct termination of incoming wide area services. It should be noted that at the moment very few manufacturers offer direct DSL termination, so the minimum expectation should be for Ethernet and 4G/LTE capability.

A key consideration here, is understand if by removing the hardware you also find carrier service levels are reduced. Note most carriers use edge termination devices to proactively monitor the connection. Removing this may leave them reactive only to service issues.

2. Traffic control and visibility

The central orchestration platform should allow dynamic distribution of traffic across multiple WAN connections based on the needs of applications as well as on rules using IP addresses and specific circuits. It must also be circuit-agnostic allowing Internet, MPLS and other flavours of circuit to be used.

3. Network segmentation

It should automate the creation of wide area segments, allowing the easy separation of business units or different application traffic types for increasing security, performance and compliance. Don’t assume all services are equal in this area – understand the use and application beyond the vendors feature name.

4. Integration of additional network services

It should support service chaining by integrating with WAN optimisation controllers, firewalls, web gateways, cloud providers and security devices and solutions.

Buyer beware

The remarkable hype surrounding SD-WAN makes it more important than ever for potential buyers to test in the most robust way possible the claims made by contending vendors. Buyers must be prepared for answers couched in terms that make-like-for-like comparison harder than it would be with a more mature technology.

The cost savings claimed for SD-WAN derive from three main areas.

  • Connectivity
  • Hardware
  • Management overheads

Will SD-WAN make MPLS redundant?

SD-WAN vendors claim that they can introduce QoS, prioritisation and
acceleration, but keep in mind that SD-WAN is an edge technology and that
the internet is a public network that lacks end-to-end control.

SD-WAN offers increased flexibility and capability, but at Gamma we are not seeing significant displacement. In the UK, MPLS is continuing to come down in cost, and many enterprises are responding by driving towards hybrid networking rather than swapping all-private for all-public networks. We see some enterprises flirt with SD-WAN, but decide to put their money into lower cost fibre Ethernet which at the moment offers a more favourable bandwidth vs cost return.

Globally, a recent Gartner survey indicates that only 20% of enterprises that have deployed SD-WAN have completely replaced MPLS.

Even in regions such as the US where MPLS is much more costly than in Europe, key enterprise applications including voice still need latency and bandwidth guarantees.

It should be noted too that the extent of SD-WAN deployment has a profound impact on forecast returns. If only a few MPLS circuits are replaced with SD-WAN, then any potential reduction in costs will be curtailed. Overall, we believe that the decision of whether to use MPLS or the Internet should not be based on cost alone. Rather it should be led by the applications being used, their criticality to the enterprise, and their particular needs. If applications are not sensitive to latency, packet loss and jitter, then maybe MPLS is not needed. However, if voice and video are being used and are critical applications, then only MPLS can provide a guaranteed business grade service.

We caution would-be buyers of SD-WAN that even in ideal circumstances it may be up to three years before savings are realised from changes in circuit type alone.

We have a toolset and the expertise to help organisations understand the likely savings and evaluate what blend of technologies is best able to support their applications.