Cloud

Demystifying Software-defined Networks Part VI: SD-WAN Adoption Accelerates as Platforms Mature

Javier Guillermo By Javier Guillermo Principal Consultant September 19, 2018

It’s been a few years since the promising new future technology called SD-WAN came on the scene, so is this related to SDN or a new concept?

SD-WAN stands for Software-defined Networking in a Wide Area Network. It shares key pillar concepts of SDN like separating the control plane from the data plane and the centralized control of the network via the SDN controller. They both allow the enablement of automation and orchestration of network devices. So, what’s the difference then? It’s like a forest from the trees expression as SDN has multiple use cases: Application Delivery Networks, Central Policy Control, Terminal Access Point Aggregation (TAP), Data Center Optimization, Virtual Core and Aggregation, SD-WAN, etc.

SD-WAN is an application (one of the many applications) of SDN technology with a focus on Wide Area Networks, allowing companies to build higher performance WANs using lower-cost internet access technologies.

SDN

Figure 1: SDN Use Cases.

The Benefits of SD-WAN

SD-WAN was designed with the idea of solving challenges like optimizing network connectivity between conventional branch offices and data centers and MPLS (Multi-Protocol Label Switching), deploying or modifying existing services in a much faster and efficient way, network congestion, packet loss, jitter, latency, etc. The reality is that “old traffic flows” are not designed for the explosion of traffic and bandwidth due to the success of cloud computing and on demand multimedia applications (think of live music, video streaming, etc.). The other major issue is not a technical one but about Operational Cost (OPEX). T1 or MPLS circuits are expensive, the former may have better point to point performance, but it is static, while the latter is highly configurable. SD-WAN technologies are aiming to bring the cost per Mbyte down by at least 60%, according to latest estimates.

Another added benefit of SD-WAN is that will work over a variety of media (for example you can also use a wireless connection), allowing service chaining, policy based centralized control, application intelligence, automation, flexibility and elasticity, etc. The reason why Internet connections weren’t used for enterprise WAN services was that the internet was always a conglomerate of different technologies best effort networks. Simply put: It wasn’t reliable or secure enough for most corporate needs. SD-WAN was designed to change all of that.

Some of you may be thinking, yes, all of that sounds fantastic Javier, but like with other implementations of SDN, you will need to make a huge investment as most solutions consist on both a central controller (often hosted in the cloud) and access nodes on-premises that support the technology, meaning you will have to throw away a lot of old equipment and make a big investment in new premises equipment, right? And how about what you mentioned at the beginning about SD-WAN being mainstream already, aren’t we really years from that?

Yes and no.

Leaders in the SD-WAN Space

Remember the blogs we wrote about the three different kinds of SDN (Open, APIs and Overlays)? While Open SDN would require a higher CAPEX investment but will bring additional innovations and advantages, SDN over overlays and SDN over APIs will be ideal for brown field development and reuse of legacy equipment. To help make SD-WAN a reality for companies, two of the leaders in this area: Cisco and VMware have made some bold moves.

Cisco bought Viptela for $610 Million and it is going to make its SD-WAN technology available not only on all ISR and ASR routers but will also on ENCS 5000 routers that are around 4 years old. That will mean in practical terms, that Cisco will push SD-WAN in over 1,000,000 routers in a question of weeks, the most massive mainstream implementation of this technology. This is great, right? Not if you’re a customer that has spent years trying to uncouple themselves from vendor lock in.  One of the key benefits for SDN implementation was to avoid closed systems, utilize inexpensive white boxes instead, avoiding vendor hegemony and lock-in again.

Cisco, like most networking manufacturers, want to keep their hardware hegemony as long as possible, for obvious reasons, and they are not shy about touting the advantages of one end-to-end Cisco SD-WAN solution.

Figure 2: Why Cisco SD-Branch is Better than a ‘White Box.’

The other leader in this space, VMware, also recently purchased (November 2017) a leader in SD-WAN technology: VeloCloud, for an estimated $449 million (according to Futuriom). Although VeloCloud offers multiple x86 appliances options with the software preloaded, it was designed to run on any x86 multi core hardware and offer some additional features like active network performance measurement (BFD), Forward error control and comes on several flavors (Premises or Cloud for Viptela and Internet, Hybrid SD-WAN or Premises for VeloCloud). Both Viptela and VeloCloud work as an overlay, support zero touch provisioning, have North bound REST and support Policy provisioning via the controller.

Although VMWare has a full SDN-NFV ecosystem with its NFV3.0 (including VIM, SDN Controller NSX, vRO for Orchestration, etc.), it is not trying to force customers into a monolithic approach.  In fact, VMware is even allowing a closer integration with Openstack thanks to VIO (VMware Integrated Openstack) and VeloCloud also works with a non-VMware ecosystem as well.

Customers will have to weigh the pros and cons of a closed system versus a vendor independent approach. If Cisco’s bet on the closed system pays off, they will be bringing back the vendor lock-in approach of the 90s, having an all end to-end-Silo from the hardware at the bottom, to the NFVI, VNFs and Orchestration.

Figure 3 – Vendor hegemony Trojan Horse? (Source: Martin Kozlowski)

Summary

SD-WAN is becoming completely mainstream but the old discussion of having open multivendor systems where the customer chooses the best for their needs versus a single vendor silo seem to be making a comeback. In total fairness, every option has pros and cons, one silo of a company could theoretically provide better end-to-end support and seamless integration between different components. On the other hand, open multivendor systems will increase innovation speed, customer freedom and speed of adoption.

Part of the SDN Blog Series

Demystifying Software-defined Networks Part V: A Decade Later, Where Are We Now? (Part II)

Demystifying Software-defined Networks Part IV: A Decade Later, Where Are We Now?

Demystifying Software-defined Networks Part III: SDN via Overlays

Demystifying Software-defined Networks Part II: SDN via APIs

Demystifying Software-defined Networks Part I: Open SDN Approach

Sources

Why Cisco SD-Branch is better than a ‘white box’

www.futuriom

www.sdxcentral.com

 

Javier Guillermo

About Javier Guillermo


Principal Consultant

Javier is a technologist with over 20 years of experience in the IT/Telecom industry with a focus on SDN/NFV, OSS/BSS, System integration, Automation, Cloud and Orchestration.

Prior to EMC, he worked at Fujitsu as Principal Planner/Architect, where he was responsible for introducing cutting-edge SDN multi-layer, and NFV application services, as well as building strategical partnerships with third party vendors. In addition, he worked at Juniper Networks Professional Services, Nokia, and Schlumberger, where he took different roles such as Customer Support Engineer, Solution Sales Manager, R&D Engineer and Group Manager at the US Technical Assistance Center.

Javier is also a health/fitness lover, loving to cook and being a certified personal fitness trainer. Lastly, he loves soccer and is an avid supporter of hometown team Real Madrid.

Read More

Share this Story
Join the Conversation

Our Team becomes stronger with every person who adds to the conversation. So please join the conversation. Comment on our posts and share!

Leave a Reply

Your email address will not be published. Required fields are marked *