Part 1: The Human Side of Automation – Key Operator Concerns
In this interview, Yuval Stein, TEOCO’s Associate Vice President of Technologies, discusses the human side of automation and how it relates to service assurance.
Q) What has spurred your interest in the human side of automation?
Yuval: My interest in this area has grown over the years as I’ve seen the challenges service providers face when implementing their digital transformation and automation projects. Plus, I’m representing TEOCO on the Autonomous Network team at the TeleManagement Forum (TM Forum) where we’re having a lot of discussions about the next steps for managing automation.
In our quest towards digital transformation, sometimes we forget that there are people involved who have been doing this work in the past and many of their tasks are now becoming automated. Any automation initiative needs to do more than just focus on the technology. It should also support the processes and the people. In fact, the technology piece is often the easiest part – it’s getting the technology, people and processes all aligned that can be the biggest challenge.
Q) How prevalent are these challenges when it comes to service assurance?
Yuval: This is a big concern – or will be a big concern – for most service providers today, where automation is typically viewed in black and white terms. There is often the belief that either a system is automated or it’s not – but there need to be more shades of gray. The point is automation needs to be an evolution.
Autonomous cars are a good example. With cars today, people are still driving them – but new automation has been introduced over the years, with brakes and alarm systems that can correct automatically. The Society of Automotive Engineers (SAE International) has created a scale of 5 autonomous driving levels, and this classification system is now used by that industry. It’s widely accepted that in the future, driving will be fully automated – but this will be a slow evolution. Each new step must be carefully tested, so that people’s confidence, along with technical capabilities, can grow over time. In a sense, telecom networks are the same, and the industry is considering establishing autonomous levels for telecom networks management.
Q) What are some of the biggest hurdles service providers are dealing with today when it comes to automation and service assurance?
Yuval: There is a lot of confusion within the industry as to what use cases should be automated first. And there are different levels of automation required, many of which can only succeed by breaking up complex processes into smaller steps. Some of these steps can and should be automated completely, but some should continue to involve human decision making, at least initially. Once these processes are divided you have a combination of automated and non-automated tasks, so it’s all quite complex. Teams then need to learn to work with this hybrid approach.
There are real challenges when it comes to deciding what should and should not (or cannot) be automated. These are extraordinarily complex relationships that need to be managed and well thought out. Service providers and their vendor partners need to design a framework that includes a plan for network automation evolution. One that will allow teams put mechanisms in place so that some steps can be automated and some not – with a realistic path for more automation going forward.
People and teams need to interact with automated tasks in different ways than they did before. This all requires a lot of planning and forethought.
Q) This all seems a bit daunting. How do you recommend service providers begin this process – and what can operators do to help their employees make this transition?
Yuval: My first recommendation would be to Maintain a Holistic View. Teams must be able to manage everything together and see it all in the same place- sort of a dashboard view. Not everything will be automated, but what is will impact other areas, so there needs to be visibility. They need the entire picture to first understand what has already happened to decide on what needs to happen next.
My second recommendation is to Avoid Data Overload. There should be a focus on showing teams information that is relevant to them. This seems contradictory to my first point- but if we think of it from an operational point of view, most teams only need information that is related to their area of responsibility, but even when decisions and actions are automated- they should still be able to review this information. Even when tasks are automated, teams need to be given access to and oversight of information that impacts their role.
Next is Stay Engaged. When it comes to managing a network, there is no such thing as flipping a switch and walking away – even with the most automated tasks. If automation impacts someone’s role, then they still need to understand what was done. What parts of the network were touched? What parameters were impacted- and why was it done?
This touches on the concept of closed loops, where it’s not just about the execution of an automated event, but also the reasoning and the decision making behind it. With machine learning algorithms, this can be challenging. Sometimes even the data scientists have trouble explaining why – but speaking as a solution provider, we should do our best.
Closed-Loops, as currently defined by various standards organizations and open source projects, are specific cases of automated, metadata-driven operational flows. These will evolve over time from being static, towards becoming more dynamic, automatically executing policies and algorithms when the need arises. The example below (Figure 1) demonstrates a hybrid approach between human teams and automation. It shows how a typical closed-loop process for resolving network bandwidth requirements also considers where human interactions and decision-making will be needed.
Figure 1 –Defining the Operational Steps of Automation and Human Decision-Making in a Closed-Loop Process Should be Carefully Considered
Lastly, I would recommend that service providers Set Smart Parameters and Policies. As more and more systems become automated, complex relationships are developing between what functions are automated and which are manual. These need to be clearly defined. Many operations are more complex and dynamic than just automated and not automated. Semi-automated is the area in between – when automatic steps are run in a more controlled manner. Where a function runs automatically- and then stops periodically so someone can validate each step along the way. Service providers need methodologies in place to cover an entire range of possibilities.
Q) What about tools and technology? Are there certain things CSPs should look for when selecting the right service assurance solution to enable automation in this step-wise fashion?
Yuval: Yes, my first recommendation is there should be a Unified Architecture to address both automated and non-automated tasks. Some network service providers believe in dividing these up, but in my view this is totally wrong. It may work for some things, but if you need to partially automate some steps you can’t because the two sides don’t speak to each other. The solution architecture should support a full range of automation and work holistically.
Another consideration is Visibility. This is mainly about the user interface – allowing people to see and understand what’s going on. With automation, many are less concerned about the User Interface, but it is still important. People still need visibility.
As we move towards more and more automated systems, data needs to be shared seamlessly. There also needs to be more information and standards for exchange of metadata so that the structure of data itself can be passed between different systems.
And lastly is Data Availability. If you need to monitor something you must understand the current state and structure. You need to have the right data available when you need it – and it must be accurate, or your conclusions will be wrong. Today, especially with network virtualization, information is very dynamic. Systems that provide data and show the current state of the network should be more dynamic and reflect the current situation. They can no longer just update the network picture once a day, as was done in the past. Now, it’s all about active inventory- because with automation, you need network information that is correct and available immediately.
TEOCO’s Helix 11 supports different levels of automation. It includes steps for managing tasks related to automated alarms, along with the ability to create automatic or semi-automatic actions. Users can select our recommended procedures – or custom create their own. Secondly, our algorithms provide unique views to the results of root cause analysis and create a better understanding of automated behaviors.
Yuval Stein, AVP Technologies