I am writing down these observations in the plane on the way home after an exhausting but rewarding week at Cisco Live 2018 in Barcelona. This is always a good show, in Barcelona for the first time after a few years in Berlin.
One of the big themes of the show was artificial intelligence and machine learning. Seemingly every other presentation referenced the transformational promise of AI and ML. Speaker after speaker stood up and referenced examples of incredible volumes of data sifted for surprising insights, or enhanced abilities to identify and serve particular requirements.
As a fan and evangelist of AI & ML, then, I worry that talking up its potential benefits without providing ready-to-use ways to achieve those results risks putting off more people than are encouraged.
However, over the course of the event, a pattern started to emerge: The presenters were not themselves offering these insights or proposing to help others achieve similar results. Instead, they were pitching the enabling tools with which interested parties could outfit their own AI & ML efforts. Faster storage, easier access to compute capacity, better distributed sensing — all relevant and valuable, but all ingredients that would have to be combined at the cost of significant effort in order to address any concrete requirement.
As a fan and evangelist of AI & ML, then, I worry that talking up its potential benefits without providing ready-to-use ways to achieve those results risks putting off more people than are encouraged. I wrote recently about the risk of getting AI projects stuck in the dreaded Trough of Disillusionment; I would like to avoid spending too much time stuck unproductively down there, and would prefer that we move on as quickly as possible to the Plateau of Productivity.
Now, admittedly, I was not able to attend all the many sessions at the show, or even a representative sample. I had to help run a booth in the Cisco Investments pavilion — where Moogsoft was recognized in Next Horizon Innovation — and deliver my own presentation.
Furthermore, I do not want this to come off as an anti-AI rant, since I suspect this would not only be pretty hypocritical, but would also be overtaken by events in pretty short order. The field of AI is developing fast, and there are new applications every day. The trick is selecting the right application, and avoiding doing AI for AI’s sake, or out of a misguided drive to keep up with one’s peers.
Choose the Right Use Case for AI & ML
Not all of us have the wherewithal, or even the need, to roll out a massive network of IoT sensors, capture everything in a data lake, and then process the take using the latest algorithms for insights. Talking about the latest facial recognition algorithms is interesting, but for most of us it’s rather abstract.
Instead, I would suggest that we focus on needs that are closer to home, and which can usefully be addressed with algorithmic and learning techniques. I’m an IT Operations guy, and as it happens, ITOps is ideally placed to take advantage of these approaches.
ITOps has Plenty of Data
IT doesn’t have to go looking for data, data comes to us. Monitoring agents already provide an endless stream of events, and modern self-instrumented infrastructure is further increasing that data flow.
ITOps Data are Ready to Use for AI & ML
We don’t have to worry about the format of the data. Monitoring events are set up from the beginning to be easily consumed by software. In fact, the problem is getting them into a state where they can be easily consumed by humans.
The ITOps Use Case is Well Understood
Finally, we are attempting to add automation and learning to processes that already exist, so we have good models to refer to when we are starting out, and to compare our results to along the way.
What was interesting to see at Cisco Live this year was that, while presentations highlighting the visionary forward-looking ambitions got all the limelight, once people had filtered out of those presentations, they seemed to be struggling to find ways to apply what they had just heard in their day-to-day lives.
Many of the pie-in-the-sky promises from the AI prophets turn out to require hiring a whole bunch of data scientists in white lab coats, and deploying all sorts of specialized software, not to mention having to compete with Bitcoin miners to get hold of enough GPUs to do something useful. All too many of these projects end up coming to nothing, or at least nothing that matches a concrete business need.
This breeds skepticism.
Moogsoft can Deliver on the Promise of AI & ML for ITOps
Moogsoft AIOps is a tool for IT Operations teams that is ready to use. Sure, behind the scenes it employs any number of data science and machine learning techniques — we’re up to 16 patents granted, and nearly twice that many going through the approval process. However, Moogsoft users are not required to become data scientists themselves. We have plenty of those in-house.
Our users, on the other hand, have spent years or decades of their careers becoming intimately familiar with specific aspects of IT infrastructure, networks, databases, or applications. We are applying AI and ML techniques to the specific use case of automating away everything that is distracting these experienced and valuable specialists from doing their jobs — ultimately providing value to the businesses that employ them.
This message of automation of routine tasks seemed to resonate well with attendees at Cisco Live, and we look forward to following up and continuing the conversation with all the existing and future users who visited our booth or attended one of our presentations.
Meanwhile, if you have been wondering where to find some substance behind the AI hype, we at Moogsoft would be happy to show you what we have built, and discuss how it might work for your IT Operations needs.
About the author Dominic Wellington
Dominic Wellington is the Director of Strategic Architecture at Moogsoft. He has been involved in IT operations for a number of years, working in fields as diverse as SecOps, cloud computing, and data center automation.