SKILup Day: AIOps MLOps Highlights
By: DevOps Institute
November 16, 2020
What a month it has been at DevOps Institute so far! To kick things off, we declared October as AIOps and MLOps month! As Artificial Intelligence and Machine Learning are rapidly adopted by the enterprise, we wanted to explore intelligent systems and help the community SKILup on all things AIOps and Machine Learning. To align with the monthly theme, we hosted several events and shared a variety of new resources dedicated to AIOps and MLOps.
Another huge announcement happened this month. From December 7-11, DevOps Institute will be hosting The Global SKILup Festival! The Global SKILup Festival is a week-long virtual event dedicated to DevOps upskilling, a deep-dive into technology trends, and career advancement. Learn more and sign up here to stay up to date leading up to the festival.
In addition to the SKILup festivities, DevOps Institute debuted a new website! Check it out here. We also continued to highlight contributions from DevOps Institute Ambassadors throughout the month of October. You can see the latest contributions on our blog, as well as our Medium publication, The Humans of DevOps.
To kick-off the AIOps and MLOps festivities, several ambassadors hosted a Global DevOps Institute Ambassador CrowdChat. The CrowdChat was an interactive social experience that included insights on key AIOps and MLOps topics such as how to define AIOps, myths and misperceptions, whether to trust AI algorithms, use cases and more:
We also live-streamed the CrowdChat backchannel on YouTube here if you want to see the video chat amongst the hosts.
On October 15, we hosted the main event – SKILup Day: AIOps & MLOps! The one-day virtual conference explored the rise, use cases, benefits, and trends of intelligent applications now and in the future. SKILup Day featured ‘how-to’ lessons from speakers: Wes Cooper, Harald Burose, Christoph Engelbert, Marco Coulter, Antonio Linari, Adam Frank, Andi Mann, Peter Waterhouse, Todd Underwood, Peter Maddison, Justin Griffin, and Garima Bajpai.
In addition to a full day of sessions, the event offered yoga, a scavenger hunt, networking lounge, exhibit hall, resource library, and even a DevOps inspired mixology class!
If you missed October’s SKILup Day dedicated to AIOps and MLOps, don’t worry! We’ve got you covered with a quick round-up of the key themes that emerged from the sessions and conversations around the importance of the topic.
You may ask, why devote a full day of learning to AIOps and MLOps?
As humans, we interact with Artificial Intelligence on a regular basis, from facial recognition security applications as we enter a store to Facebook algorithms that feed us content. As it relates to the Humans of DevOps, machine learning and artificial intelligence are rapidly being embedded into more enterprise applications, giving way to practice areas in MLOps and AIOps.
The DevOps Institute SKILup Day explored the rise, use cases, benefits, and trends of intelligent applications now and in the future. SKILup Day speakers explored a variety of themes around AIOps and MLOps including, defining AIOps and MLOps, adoption and maturity, the future of intelligent applications, lessons learned, and the humanity in AIOps. Below we look at key quotes and discussion points from the day.
What is AIOps?
Throughout SKILup Day, many presenters discussed the meaning and principles of AIOps and MLOps. Andi Mann from Splunk’s session titled,” Using Data Analytics, Machine learning, and AI in IT Operations” took a deep dive into AIOps. He began by sharing Gartner’s definition of AIOps: “AIOps combines big data and machine learning to automate IT operations processes, including event correlation anomaly detection and casualty determination.” Mann explained AIOps further, “This is about using advanced techniques for data analytics. So going beyond just spreadsheets and straight-line productions to really use analytics to understand what is happening inside a system.”
Wes Cooper from MicroFocus also explored what AIOps is during his session, “Taking DevOps to the Next Level with AIOps.” He added, “An AIOps solution has to be built on the premise of machine learning and big data analytics and being able to leverage those things to help augment some of the processes for operators and help them to get a better understanding of the environment.”
How do you adapt and mature AIOps/MLOps applications?
Many presenters offered sessions that discussed “how-to” adopt and mature AIOps and MLOps. Sessions included varying insights depending on the presenters’ expertise. Marco Coulter from AppDynamics, for example, presented a session titled, “The Stepping Stones of AIOps.” He sets up a realistic expectation for every AI adoption journey, “You’re not going to reach full AIOps functionality overnight. You’re going to mature through stages.”
Peter Maddison from Xodiac presented, “Deciphering AIOps and MLOps,” which explored the similarities and differences between each. Maddison also shared the importance of avoiding automation silos, “We need to make sure that we’re standardizing the system as well so that we’re consistent in how it’s implemented so that we’re not ending up with silos of automation which can’t talk to each other.”
During the session, “Taming Data Science Dragons with MLOps and Kubernetes,” Peter Waterhouse from RedHat talked about the challenges of making a business impact with AI investments and how to leverage systems like Kubernetes to help drive that business value. Waterhouse said,” It’s really about harnessing today’s technology that’s actually making data science, AI, and machine learning possible.”
Antonio Linari from Expert.ai supported this concept of leveraging today’s technology for AIOps and MLOps by offering a demo of a semantic search engine using a small edge cluster during his session, “Cloud Native on the Edge.” During his session, he also talked about growing power requirements, “Today computation, especially related to deep learning, takes a lot of power…consumption became 300,000xs more from 2012-2018.”
Remember, you can learn from outages
Speakers also highlighted the challenges and inevitable outages associated with AIOps and MLOps. Todd Underwood and Daniel Papasian from Google examined machine learning outages over the past 15 years, focusing on the importance of understanding failures and never letting them go to waste. During the session, Underwood stated just how common outages are, “ML systems break all the time.” Further, he added, “We often don’t talk about our failures, but the failures are where interesting things happen.”
Justin Griffin from RigD focused his session on effective incident response. Griffin shared details about the substantial cost of downtime time and highlighted the opportunity for organizations to improve their incident response. He asked an important question, “When you think of CI/CD, why is continuous process improvement not also a foundational element of incident response?”
AI-driven Observability is important
A theme that was repeatedly covered was AI-driven observability. Adam Frank from Moogsoft said, “I really believe that intelligent observability is uniquely positioned to provide that continuous innovation that we need to deliver [to our users].” He also argued that without AI, observability creates a lot of noise, “Observability, now that you are admitting all this information, without actually applying AI to it, then there’s not a lot of context to it.”
Christoph Engelbert of Instana reiterated this during his session, “Road to Observability: Centric or Agnostic – That is the Question.” He shared that as humans, we need patterns. We look at our data and try to make sense of it, but sometimes it looks like gibberish. “We need a machine to take all our different data sources, to put them together and correlate them,” said Engelbert.
Don’t forget the human side of AIOps and MLOps
In true DevOps Institute form, Helen Beal explored the humanity of AIOps and MLOps. During her session, “Surveillance, Capitalism or Bust” she explored the commodification of personal data and how “We’ve become the product.”
She explored the advantages and disadvantages of surveillance capitalism and explored the idea that “AIOps is humane. AIOps helps us be humane.” Participants in the live panel featuring Andi Mann, Chris Engelbert, Garima Bajpai, Jayne Groll, Marco Coulter, and Todd Underwood also explored the human side of AIOps and MLOps, concluding that humans need to understand the amount of hard work that goes into these intelligent applications. Panelists generally agreed that a lot of planning and experimentation needs to happen for AIOps and MLOps to be successful.
What is the future of intelligent applications?
Garima Bajpai explored this during her session, “AI as a Product.” She shared that, “We are moving in a direction where personalized, customized digital solutions are taking precedence. We need to ensure that we have ways to use data sources.”
Want to know more about the sessions? For a quick recap of each, check out the sketches below (add graphics). You can also view the videos and download the slide decks by viewing the DevSecOps SKILup Day on-demand.
There are plenty of events, fresh content, and exciting announcements in the pipeline. We’ve designated November as Continuous Testing month! Stay tuned for details about our next Global Ambassador CrowdChat and be sure to join the next Virtual SKILup Day on November 19.
Also, be sure to complete the Upskilling 2021: Enterprise DevOps Skills Survey for a chance to win a Nintendo Switch!.