March 14, 2017
Believe it or not, the first known reference of interoperability testing goes back centuries to the days of Homer and his Iliad. Since then the concept has taken full measure across all aspects of civilization and society including business, law, engineering, medicine, sciences, social sciences, etc.
Interoperability testing and testing events (aka “plugins” or “plugtests”) do not go back as far but have proved immeasurably important in advancing industry-wide collaboration ultimately for the benefit of end users. ETSI, the organization responsible for overseeing the standardization of NFV, itself has been hosting plugin events since 1999 for the expressed purposes of improving standards and reducing product time-to-market.
Most recently, ETSI hosted its first NFV Plugtest involving both commercial and open source technologies/products across three categories: Virtual Network Functions (VNFs), Management and Orchestration (MANO) solutions, and NFV Infrastructure (NFVI) with pre-integrated Virtual Infrastructure Manager (VIM).
Over the course of almost two weeks, there were 146 testing sessions each a grueling three hours long where any company or organization could volunteer its wares to assess the interoperability level with other products and to ensure the correct interpretation according to ETSI NFV specifications. Of these, RIFT.io was involved in more than 50 sessions testing both our commercial solution as well as committing engineers to help test OSM implementations.
The testing revealed some notable results including:
- The industry itself proved to be in a far more interoperable state than expected. Going into the plugtest, ETSI expected the rate of interoperability success to be around 30 percent. The testing netted results close to 98 percent, albeit some of the tests had to be more “engineered” than others to achieve interoperability. In general though, we can determine that the vendor community behind the different aspects of NFV to be following ETSI specifications pretty closely.
- The importance of standard, data model-based descriptors for configuration could not be emphasized strongly enough. Irrespective of descriptor modeling technology, YANG or TOSCA, using common descriptors is essential for efficient network operations and onboarding of new network services. A pleasant surprise for us was the wide familiarity among the plugtest participants of RIFT.io descriptors, which we consider to be a competitive advantage for automating network service deployment and onboarding new VNFs.
- Cloud-native products and technologies fared far better and were more relevant for operator applications than those migrated from IT data centers or appliance-based tools. This was certainly the case for VNFs that we tested where we saw that network scaling-in was not baked into their base models. Also, the concept of network services not well thought out in many of these options in terms of delivering an end-to-end lifecycle management of services.
Of course, these are areas where RIFT.io excels. Our RIFT.ware platform is a fully ETSI-compliant NFV MANO solution that was born to simplify the deployment of multi-vendor VNFs and orchestration of complex, multi-vendor network services in telco and enterprise clouds. So while many of our fellow participants worked hard to ensure interoperability, we were largely there, meeting or exceeding our expectations going into the plugtest.
Perhaps the most important lesson that all the participants learned is that there’s a long way to go to get all of the key NFV elements truly carrier-grade. Given the significant progress we’ve made so far, we can say confidently that this is only a matter of time. Thus, we look forward to the next plugtest where the industry as a whole can measure progress against the benchmark we established in the first test. More importantly, we look forward to sharing progress reports from actual product deployments in operator networks. Stay tuned for those updates.