New Research from Appledore Research Group plots the status and trajectory of the industry and leading players.
Appledore Research Group has recently released the 2nd major report in a series of 3 that investigates the necessary foundation to achieve flexible, efficient closed loop automation in virtualized networks. This report is available here. Appledore Research Best Practices Part 2 report
The report identifies 4 key technology enablers; in this blog I will comment on two that are closely linked: First, Model- & Policy- driven orchestration, and Second, onboarding tools and practices to help Service Providers build, maintain and share those models and rules.
Without diving into the details, we wanted to emphasize that it is essential that orchestration be effected natively by models; that those models must reflect intent; and that rules (or other methods) must be capable of instantiating that intent based on context. Any old model won’t do.
The good news is that nearly every orchestration vendor is delivering model based orchestration that utilizes rules to specify at least some parts of intent. But it’s also clear that most are still leagues behind the “HyperScale” Web leaders (Azure, AWS, GOOG, BlueMix) in terms of true hands-off automation, and fully intent-based, autonomous operation. That is why Microsoft’s Azure, despite being deployed in adjacent industries, is shown as the most evolved single system in our industry progress chart. Microsoft documented and defended both the method and the evidence of truly intent-based, autonomic operation, at huge scale. For the most part, telecom examples have been less intent based, vastly lower scale, and aimed at very limited use cases.
In the core of telecom orchestration we are seeing progress. Major players have evolved on our stated axes, including Nokia’s CloudBand, Ericsson’s Cloud Manager / Dynamic Orchestration, and Netcracker’s 10. More importantly all of these claims are supported by major Service Provider deployments operating real, cash-generating services, at meaningful scale. Others have documented less; HPE, Oracle and Huawei provided relatively less information on their ambition for automation, implementation specifics, and deployments and metrics achieved. Cloudify, a disruptive entrant, demonstrates a very clear ambition to implement clean model-native, generic orchestration. (Watch for upcoming research on what intent encompasses and how to best achieve it).
From conversations with suppliers, as well as with Service Providers, we believe that the modest pace of observed change is less a result of suppliers’ capabilities, and more one of industry ambition: many suppliers indicated that their Service Provider customers – who determine what is actually delivered after all – are hesitant to fully turn over control to “The Machines”. We are also told that those SPs are focused on implementing very specific service and automation tasks, rather than designing orchestration methods that are truly generic (and who’s operation is entirely determined by the model and context). If the goal is to implement something well understood and relatively simple, sometimes its easiest (and cheapest, and quickest…initially) to simply create the logic rather than define a universal, model-driven mechanism. And from what we observe, some of that occurs.
We encourage anyone interested to read the series of three framework research reports that outline:
- What is the state of “MANO+ orchestration” as of 4Q2017?
- What is the ambition, status and direction of 20+ leading players?
- What are the best practices to achieve true automation?
The agility and profitability of our industry depend on success.