Public vs. Private: Battle for the Future of SD-WAN

The Internet is perhaps the greatest of mankind’s infrastructure achievements. It has virtually unlimited use cases. For Enterprises, it has become an integral part of connectivity, with over 90% of SaaS access leveraging best-efforts public Internet. In fact, many SD-WAN deployments rely exclusively on Internet for connectivity.

There are a growing number of Enterprise use cases, however, that have exposed the inherent limitations of these public networks: applications that rely on real-time protocols, like video and voice; modern SaaS and IaaS, particularly over long-haul distances, across particular geographies, or at certain times; large file management; and connectivity that requires the highest order of security for facilitated compliance.

There is an anecdotal feeling of Internet insufficiency or unpredictability that has led the market to offer various Internet enhancements or alternatives. SD-WAN vendors, for example, support Internet last-mile provider redundancy to achieve high standards of performance. Beyond the last mile, a variety of Internet core alternatives have come to market – some built around optimizations or enhancements of the public Internet backbone, others eschewing public IP networks for software-controlled private cores.

MPLS is still the gold standard of Enterprise connectivity, even though it is itself far from perfect in its performance and privacy dimensions. Still, it is very unlikely that Optimized Internet solutions can approach the SLA-backed MPLS standard for performance, despite frequent claims to the contrary. These approaches use a public Internet core they don’t own – an ultimate black box removed from discrete visibility and control. POPs can probe this black box for best routes, but the quality of measurement is gross, and the reaction speed to fast network changes, slow. These outside-in techniques fall short of what true software-based measurement and control of an accessible core can deliver. WAN optimization techniques are best layered on top of a wholly software-controlled core, but on their own they cannot deliver MPLS or better performance.

Beyond Optimized Internet, Cloud MPLS is the next, logical Internet Core alternative. These solutions add POP-access flexibility to traditional MPLS networks. Unfortunately, these designs retain the high cost of MPLS. In addition, they require decryption of Enterprise traffic at the POP level, an unacceptable requirement for businesses committed to strict security and facilitated compliance.

Beyond these approaches lies the ultimate Internet Core alternative: a true, autonomously controlled software-defined private core.

There are two categories of prospective customer for this newcomer among SD-WAN connectivity options.

Enterprises currently using MPLS for mission-critical connectivity are well aware of the high costs of these private networks, compounded by exploding bandwidth use. As they turn to software-defined networks as cost-effective alternatives, they must be assured of performance equivalency with MPLS.

On the other hand, businesses who have contributed to the rapid growth of software-defined edge technologies like SD-WAN, and clouds applications generally, have largely relied on public Internet for primary connectivity. When these companies experience the downstream effects on Internet unpredictability – glitching real-time applications (e.g. videoconferencing and VoIP), inconsistent long-haul traffic, poor SaaS and IaaS performance, or compromised security and compliance – they need to know that the global connectivity alternatives under consideration can actually outperform the Internet for those use cases.

Designing a meaningful study of public versus private network performance in the context of the modern hybrid Enterprise poses significant challenges. The unpredictable nature of the Internet, combined with its massive scale and rapidly-changing peering structure, complicates quantitative comparison. Measure too little for too short a period of time, and the data may be anomalous.

Mode and its service provider partners have, over time, collected 320+ million data points, among 32 last-mile locations, 24 cloud instances, two cloud providers, and four continents – to produce a study of substantial power. The results of this extensive research will be published shortly as The 2019 Public vs. Private WAN Performance Comparison.

The goal of the comparison was to measure the use of public Internet vs. a Mode Core’s archetypal private network in some of the most common Enterprise connectivity cases, with enough sample breadth and duration to produce a study of substantial power. This Comparison was aimed at businesses who have relied on SD-WAN over Internet to date, and now seek an economical, SLA-backed high-performance Internet alternative. The results provide a definitive answer to the question: does a private network (e.g. Mode Core) offer meaningful performance benefits over best-efforts public Internet for some of today’s most-vital Enterprise use cases? For the answer, please visit and be among the first to gain access.


Author's Bio

Paul Dawes

CEO at Mode