Analysts are not employed by SDxCentral and the views, thoughts, and opinions expressed in their content belong solely to the author and do not reflect the views of SDxCentral. Note: AvidThink is a separate organization, created by Roy Chua, that is not affiliated with SDxCentral.
In the run-up to the upcoming trade show season automation is dominating discussion in the telecommunications network software domain. However, as it usually goes in early technology adoption cycles, the first practical experiences with automation have opened up a number of questions in three areas:
- What are the practical uses of automation?
- What capabilities does a solution needs to be considered an automation solution?
- What should be automated?
The primary motivation for automation in telecommunications networks is to solve two overarching problems:
- Complexity: The growing number of clients, services provided, and increased traffic is making telecom networks more complicated. As the number of domains and network elements increase, so does the network management workload. A lot of operators have realized that scaling their network-related workforce with the growth of their networks is untenable — not only due to increasing costs, but also due to the limits they’re facing in the labor market. Simply put, not only is hiring the required number of network engineers costly, but in many cases the skills required are not available in the workforce.
- Human limitations: Cloud services, network multi-tenancy, and competitive pressures from within the industry and beyond require telcos to provide their services in a more agile way, cutting down service provisioning time to seconds or minutes, rather than days or weeks. The number of network management operations in such an environment quickly grows, and the “time budget” for each operation declines. Humans can’t cope with the operational tempo required.
The industry has recognized that some form of automation is the most logical way of solving complexity and human limitation problems. But, the actual best practices of automation are still being developed, and the limits of automation have yet to be defined.
The Limits of Automation
Operators should consider two types of limits when deciding on their automation strategy:
- Usefulness: Operators should initially limit their network automation initiatives to use cases that bring palpable business benefits. Automated network element provisioning and repetitive processes that tie human resources can, and should, be fully and truly automated today. Service provisioning for both retail and wholesale customers is also easily automated. Both bring significant cost savings and agility gains, and the underpinning technologies are already being deployed. Automated service provisioning also brings operators closer to the goal of opening up their infrastructures to their enterprise clients and other operators in a controlled way, enabling operators to provide their networks to clients “as-a-service.”
- Technology limitations: Machine learning and data analytics are receiving a lot of attention, but finding actual deployed telco use cases that rely on these technologies is very rare. Additionally, artificial intelligence (AI)-driven closed loop mode of operation in telecom networks introduces a degree of unpredictability into the network. With an increased emphasis on deterministic links (for IoT, for example) and stringent QoS levels for enterprise services, vendors will have to make any AI-driven system adhere to these principles. Furthermore, from a purely financial standpoint, it is questionable whether replacing small numbers of human engineers involved in processes heavily geared toward decision-making (like network planning) will make much sense.
Automation Should Go Beyond Empowering the Engineer
For all the focus on automation, it is too often viewed as synonymous with the evolution of traditional network management systems and tools. Much of the discussion around automation does not move far past comparing it with scripting — a tool that network engineers have successfully used for a number of years. If viewed superficially, automated mobile backhaul provisioning, for example, achieves the same end goal as manual scripting. After all, with scripting, the engineer workload required is just a couple of clicks per cell site router. However, when network engineers are required to deploy thousands of backhaul links per day in very large networks, they run out of the number of clicks they can physically perform during their working hours. In this case, cell site routers need to be provisioned automatically — without any human intervention.
For this use case, and others like it, streamlining network management and “empowering the engineer” is not enough. Automated provisioning is only going to become more important, as concepts such as network slicing become commonplace with the upcoming 5G deployments. Similarly, automated regular maintenance (software patching and upgrades, for example) can also bring very significant increases in efficiency and network availability. For these use cases, the best analogy of the future network mode of operation is airplane autopilot — modern autopilots require very little human intervention to perform a flight and usually perform better than humans in certain situations. But humans retain key decision-making capabilities in the system and can take over in emergencies.
Another consequence of viewing automation in the light of the autopilot analogy is that the usefulness of linking automation and AI quickly becomes questionable. Modern autopilots do not require AI in any shape or form to safely and autonomously carry us through the skies. Similarly, the use cases for automation that make the greatest impact to telco operations, like automated provisioning, don’t require AI (or, to use a more precise term, machine learning). What they do require — same as autopilots — is ways to gather operational data (real-time streaming telemetry), and control mechanisms to effect changes (SDN).
That is not to say that AI has no place in network automation. But the use cases where AI can help — like predictive maintenance, long-term network optimization, network planning, and deployment automation — are currently not the choke points in network management and operations and do not seem to represent such a headache to operators as network element provisioning and management. It is worth noting that webscale firms, as pioneers in network automation, have happily automated their data center networks for a decade now, with no help from AI whatsoever.
As automation solutions become commonplace, telcos should realistically assess the value of automation in their particular network environments, and quickly establish a connection between automation solutions and business outcomes. At the same time, they should avoid overcomplicating automation solutions with AI, instead clearly defining their management processes with a view of using automation and humans where appropriate.
ANALYST COLUMN DISCLAIMER
Statements and opinions expressed in articles, reviews and other materials herein are those of the authors; not the editors and publishers.
While every care has been taken in the selection of this information and reasonable attempts are made to present up-to-date and accurate information, SDxCentral, LLC cannot guarantee that inaccuracies will not occur. SDxCentral will not be held responsible for any claim, loss, damage or inconvenience caused as a result of any information within this site, or any information accessed through this site.
The content of any third party web site which you link to from the SDxCentral site are entirely out of the control of SDxCentral, and you proceed at your own risk. These links are provided purely for your convenience. They do not imply SDxCentral’s endorsement or association. The copyright and any other intellectual property right any third party content belongs to the author and/or other applicable third party.