With the SD-WAN market in full swing, we’re seeing early signs of some maturity in the market. While early SD-WAN childhood efforts were around deployability and getting basic functionality working, I’m now seeing SD-WAN hit puberty as characterized by industry efforts around interoperability and testing of SD-WAN for scalability and performance.

Most SD-WAN implementations today are standalone, each is their own island. Many enterprise deployments today are single vendor, and by-and-large many of the early communications service provider (CSP) deployments are the same, with distinct SD-WAN domains controlled by a single vendor platform. Just a few weeks ago, if someone brought up SD-WAN interoperability, I would have said that it was premature.
However, at the recent SD-WAN Summit 2018 in Paris, I talked to a handful of top-tier CSPs that are pushing for interoperability efforts to start sooner rather than later. That, coupled with conversations from a SD-WAN Summit panel I moderated on interoperability, has modified my point of view.

Almost everyone (enterprises, CSPs, standards bodies, vendors) I spoke with acknowledges that data-plane interoperability in SD-WAN is unlikely in the near future. Enterprises and CSPs are telling me they don’t need it yet. This brings back memories of the old IPsec interoperability wars. Trying to create interoperable meshes of nodes from disparate vendors today is really putting the cart before the horse because we need to start from the control plane.

CSPs that are in the process of building or customizing their orchestration systems to integrate with SD-WAN offerings say that having interoperability at the control and management level allows for coordination between multiple SD-WAN domains. It also makes switching vendors (or adding new SD-WAN offerings in the future) feasible with much less pain.

The present efforts focus on interoperability at the northbound API level. They govern the APIs used to provision and control SD-WAN deployments. I know of at least two groups working on this effort: ONUG with its Open SD-WAN Exchange (OSE) and the MEF’s effort on using the LSO (Lifecycle Service Orchestration) Presto API as a standard northbound. I’m sure the MEF will have more to share at its upcoming MEF 2018 conference at the end of October in Los Angeles.

As we hit adolescence in the SD-WAN market, it stands to reason that we’ll see some chest-beating in performance and scale comparisons between competing SD-WAN vendors. EANTC, a well-known European test lab, announced new SD-WAN scalability test results to the 500-plus attendees at the SD-WAN Summit and they showed some impressive numbers from Juniper Networks and Nuage Networks (Nokia).

EANTC’s Managing Director Carsten Rossenhövel was quick to point out that the two tests EANTC had performed, along with the testing of Huawei’s gear last year, focused on different aspects of performance. Unlike NSS Lab’s tests (as discussed in last week’s article) which applied the same test suite to all vendors, EANTC worked with each vendor to pick an aspect to highlight and tested that aspect, avoiding a head-to-head comparison between the different vendors.

I’m mixed on this approach. On the one hand, having a consistent set of tests that apply to all vendors provides a level playing field on which to evaluate all the parties. And yet, each set of enterprises will likely have their own specific needs, and value different capabilities of SD-WAN vendors quite differently, so a single test across the board might not provide sufficient insight.

NSS Labs tests were focused more on data plane testing (this is not surprising because NSS Labs has a history of firewall and security device testing), whereas the EANTC tests concentrated on a mix of control plane and data plane, showing quite impressive numbers across all the vendor tests for Huawei, Juniper (Contrail), and Nuage Networks.

Regardless of your personal point of view on which tests make the most sense to an enterprise or a service provider, these tests demonstrate significant scalability along some key critical metrics. I am interested to see how these same platforms stand up in the real-world, outside the test labs and if they will fare equally favorably. Carsten and his team have done a good job in attempting to simulate real-world conditions, like creating intermittent failures upstream/downstream instead of outright pulling cables to simulate link or path failures. But as we all know, the real-world is hard to predict, and it’s time for us to let our bunch of teen-aged SD-WAN solutions into the world and see how they fare.