PurePerformance - AIOps done right with Wolfgang Beer

Episode Date: January 30, 2019

Wolfgang Beer talks about how to make the most of our latest innovations and enable you to automate and manage your operations at web-scale. He shares insights from his session on management zones, API integrations, deployment automation, and best practices using our open AI.

Coming to you from Dynatrace Perform in Las Vegas, it's Pure Performance! Hello from Dynatrace Perform 2019 in Las Vegas. I'm Andy Grabner and this is Up Close and Personal with Product Management on Pure Performance. I want to introduce my guest Wolfgang Baer. Hey Wolfgang, you just presented here at Perform at one of the breakouts. I just grabbed you coming out of your breakout room. What have you been talking about and who was your counterpart, who was your co-speaker and what have people missed? Hey Andy, nice to meet you here in Vegas.

I just had the session about AI ops done right together with Eric Lensness where we presented the new capabilities in terms of AI operations that we introduced lately in the Dynatrace Davis AI engine. Cool, so Dynatrace Davis, I think that's also maybe something new for folks that have been following Dynatrace Davis AI engine. Cool. So Dynatrace Davis, I think that's also maybe something new for folks that have been following Dynatrace for a while. Davis used to be associated with the voice and the chat interface of Dynatrace, and now Davis is the name for the AI engine

that you have obviously, from a product management perspective, have been driving over the last couple of years. Yeah, Davis actually has become the intelligence within Dynatrace that you can talk with, interact with, and Davis obviously helps you with analyzing all your abnormal situations that you encounter in your production environments. And what we introduced in our breakout session today was the openness that we came up with lately, where you can introduce third-party events within the problem correlation, as well as data ingest in terms of metrics, bring your own metrics, bring your own events and Davis will analyze it for you. That's pretty cool.

So let's talk about these two things. You just said external events. So give me some examples of what people are doing there. What are external events? Well as I know that one of your favorite topics is shift left and pipeline, unbreakable pipeline. One of those event types obviously is the information sent to Davis about ongoing type of problem, event, like an availability event when something goes down, or as Eric presented today, SNMP monitoring tool

that you have probably running in your environment. So that means, like this example I just said, so Eric showed SNMP, that means a lot of our customers obviously, and I'm sure that is true for many organizations around the world, they have some type of monitoring tools already that are covering certain aspects of the infrastructure, they already have their alerts, and we can now feed these alerts into the Dynatrace AI engine, so that if Dynatrace detects a problem we have all of these external events also correlated. Exactly, so when we are talking with with our Dynatrace users we often find out that large enterprises and large organizations operate a huge amount of third-party, probably legacy tools that they would like to

consolidate with Dynatrace. And so many departments also means a lot of different tools where you would not like to lose that information and Davis can collect those information through the API and will present you the correct root cause in terms of abnormal situations. That's pretty cool. So that was one of the aspects. The other aspect I think you said is extending the AI through custom metrics. I think you call it the bring your own metrics. How does that work? What are the ways to get your own metrics from other tools into Dynatrace?

Well, basically you're again using several sources or several channels where you can feed in custom metrics. Like one of the most popular ones is to write a one agent plugin or a remote plugin. But you can also use the Dynatrace API to send in your own metrics. No matter where they are coming from but I think the most interesting part is not sending in the custom metrics but that Dynatrace is capable of detecting unhealthy metrics without baselining and without thresholds at all. Carlos, I think that's a recent announcement. I remember your blog from, I think, maybe November, December, where you announced the beta availability of this enhanced AI feature.

So that means you are now, or the AI Davis is now looking at all the metrics and is figuring out which metrics behave abnormal in a certain time frame where we detected a problem. Is this how it works? Yes. So in terms of, of course, we are using baselining to trigger and to learn about abnormal situations. But in terms of the root cause analysis,

we are analyzing all the metrics. The one agent automatically detects and monitors which means by average i would say 500 metrics per host depending on out of the box depending on the technology that is running on that host but as i said before you're open to send in and to contribute your own metrics. And those or your own metrics are then first class customers or contributors to the root cause information here. That's pretty cool. So if I envision the problem screen now, I think when I've seen it on the problem ticket screen, I can see the impact that I always saw, the impact that services or applications, but then also on the root cause section, I see what we have detected as the root cause and also all the metrics that have abnormal behavior.

And coming back to the first thing, also all the events that are kind of correlating to the components that are part of this problem. So like a deployment event, a load test started, an SNMP trap, all these things now all consolidated in a single problem ticket. That's pretty cool. Yes, exactly. So everything that contributes to the root cause, no matter if it's an external load testing event or configuration change event or your custom metric behaving in a strange way. Everything summed up in the root cause section. That's pretty cool.

Hey, is there anything else that you covered in the session that you want to highlight? Things that, you know, the sessions are recorded, I believe, so people can watch the recording afterwards. Is there anything you said, this is why you should watch the recording? On top of what you already said, I know it's already a lot of content, but maybe anything that comes to mind or anything that you as a product manager want people to know? Yeah, I think the last thing to mention is, and one of the most important things to mention is, the toggle for opting in into the new enhanced AIOps features is right in the product as of today.

So don't hesitate. Toggle it on and you're fine with your root cause section. That's perfect. So you just go to the problem screen. You see it on top with a big green bar, I believe. Yes. And just turn it on. Yeah.

And be amazed. Be amazed. Yeah, it's awesome. bar i believe yes and just turn it on yeah and be amazed be amazed and uh um because you said you know we've we've renamed uh we use another word davis for the whole ai uh that means obviously the davis as we knew it like the voice ops and chat ops and the integration with alexa and and the other uh tools out there they obviously also benefit from that because the root cause detection delivers more data also for these channels, right? Absolutely, yes. Cool.

All right. Hey, Wolfgang, I think this was amazing. Thank you for that. Thank you for driving the whole AI theme. I know there's more people involved, but you are obviously one of the product managers here that are driving this. Well thanks and for Pure Performance, I'm Andy Gradner. but you are obviously one of the product managers here that are driving this. Well, thanks. And for Pure Performance, I'm Andy Gradner.