Performance, Elevated: How Instana is Advancing Observability

[00:00:00] Chapter 1: Introduction and Background

Mirko Novakovic: Hello everybody. My name is Mirko Novakovic. I am co-founder and CEO of Dash0. And welcome to Code RED code because we are talking about code and Rad stands for requests, errors and Duration the Core Metrics of Observability. On this podcast, you will hear from leaders around our industry about what they are building, what's next in observability, and what you can do today to avoid your next outage. Today my guest is Bill Lobig. Bill is Vice president at IBM and responsible for the automation portfolio, which includes a company that I initially founded, Instana, but also companies like Turbonomic. And I am super happy to have you here. Bill, welcome to Code RED.

Bill Lobig: Hey Mirko, thanks for having me. It's been a long time. It's exciting to get together again.

Mirko Novakovic: Yeah, it's four years. I thought about it feels like a year to me, but it's. It's already four years. But I always start the conversation with my first question about a Code RED moment. So did you have a Code RED moment in your career?

[00:01:05] Chapter 2: Code RED Moments and Log4j2 Vulnerability

Bill Lobig: Yeah, probably more than a few. I was thinking about this and, you know, making it very relevant to the audience and the topic here today. And the one that comes to mind is, I think, is one that everyone can associate with is a few years ago, we had this log4j2 vulnerability. Everybody remembers that, right. And, you know, and I reflected on that recently. And, you know, the answer everyone was looking for was fix everything everywhere and every version, you know, tomorrow, today. And I think largely speaking, it took many companies months to get through that. Right. Months. And it just so happened to occur on the Christmas holiday. And, you know, this is a great example of where, as we've evolved that I think AI and observability and the kind of new ways of thinking about this stuff can really help drive better insight into in that example, prioritization of do we fix the right vulnerabilities in the right sequence for the right reason? Or they attack surface area, they back office, you know, all of these things. And I think it was largely chaos. And it doesn't need to be that way. So that was what I thought about.

[00:02:12] Chapter 3: Automation and AI in Observability

Mirko Novakovic: Yeah, that's a good story. And I remember that very well, especially because I was a Java developer and I used Log4j2 everywhere. Right. So that was a pretty tough one. But talking about this, I had a different question. But that's a very interesting topic because I know that and I read that IBM is doing a lot with Watson and AI and Watson X, I think it's it's now. Right. And you also announced a lot of new features for Instana, which I always follow. And some of them are automation features. I mean, I know there is this auto discovery. So we basically with Instana, we built an agent that auto discovers all the technology, including libraries. Right. We could figure out which libraries you're running inside of your code. So we easily could say this is the code we're log for j is running out of the box. Do you have some automation on top? How is it working today? How are you using AI to automate exactly what you were just explaining? So how? How could you use Instana and IBM automation products today to solve those issues?

Bill Lobig: One of the things that we've been very focused on is so you mentioned the auto discovery, but helping people find that needle in the haystack or improve that signal to noise ratio more easily than they ever have. And it's funny, a couple years ago, you know, when IBM acquired Instana, AIOps was like this hot topic, right? And I've always thought that term was a little it's like, what does it even mean? Right? It's like I think of AI as like I like to use analogies. It's like electricity. It's like in the walls. It's all around you. You don't think about it. You take it for granted. Alone, it doesn't really do much, but it makes your house. What makes your TV work, your microwave like it powers things. So in that way, it's a feature of other things. And that's how we're leveraging it, which is a feature of observability. So issue summarization, probable root cause detection we're experimenting and playing around with. And this would be a great thing for your audience to give us feedback on, you know, automated remediation in terms of generating scripts, Runbooks, playbooks. And I personally think I'm observing people aren't quite ready for that yet. Like they want the suggestion, not the full, you know, like there's still a human in the loop there. But, you know, it's really about leveraging AI to do what it's always done, which is process massive amounts of data better than humans can, and give you a recommendation on what you should do.

Mirko Novakovic: When we started with Instana, we had this root cause analysis, right, which worked pretty good. But you have false positives. And my experience was, if you are right, 99 out of 100 times people will not appreciate if you do this one mistake, right? Especially if you wake up in the middle of the night and then you figure out it's a false positive, which probably always happens. And so one of the things I read about, and I like the idea of Dion Almar written that who is now on an AI company who was at Google, he compares it to Google Search. And he says most of the root cause analysis tools or tools want to show you the one root cause, right? Yes. Where if you search in Google, you get a list of possible results. And if the first one is not correct, you still have this second or third, which means if the third is correct, you don't care that the first one was not correct, right? Yes. And I kind of like the idea to have this kind of system where you don't say, hey, here is the root cause, but here's a list of prioritized possible root causes. And then if the first one is not correct, you still have the second or third choice. And I like the idea, which is then hard to do automation because you don't know upfront, but you can still have a user in between, which is what you are saying. It's exactly my experience that users are probably not ready yet for full automation, but that way they could say, yeah, the third one is correct and now I want to automate.

[00:06:03] Chapter 4: Evolution and Integration of Instana

Bill Lobig: I completely agree, and I like this analogy with search. And this also brings that level of explainability or transparency. So unlike Google where, you know, search engine optimization you're paying to get to the top. And here we can say here are the probable root causes and why and help people understand, you know, the reason for that. And then they can choose for themselves which one they think is the actual root cause. And to your point on, like the automations, known problems to known solutions seem like an obvious thing. But just because there's a known solution to a known problem doesn't mean you in the moment know that solution, right? So there's a lot that can be done with AI around curating that and giving those actions to the probable root causes that you may pick in the list. For example, just another example of AI is applicability.

Mirko Novakovic: Yeah. But coming back to my first question I wanted to ask because I'm really curious. I mean, I'm the co-founder of Instana, and I built that product for five years, and now it's your baby, right? So I would be very interesting what happened in the past four years. What are the exciting new things? How is Instana evolving that that would be really interesting for me to understand.

Bill Lobig: Yeah, it's been a fun ride. So when we acquired Instana, as I think you'd agree you know, it was the best and I think is continues to be the best at Cloud Native. Right. It was built for Kubernetes on Kubernetes. And that's that's really been at the core of it. But the observability market has evolved and the expectations have gone up around other capabilities that, you know, cost management, synthetics, real user monitoring, digital experience, all of these things. And truth be told, Instana lacked a lot of those back then. So we filled in all of these gaps. We've got synthetic monitoring now, remote points of presence for latency testing. We've got full logging capabilities, not just in the context of observability, but even vulnerability and some security capabilities. You wouldn't be a serious observability vendor if you didn't say you had OpenTelemetry, which we've integrated as a first class citizen. And frankly, I'm really excited about that because, you know, it avoids vendor lock in. That's what a lot of folks are interested in. But it really to your point earlier on I pushes the expectation up. We're like agents are now commoditized okay I'm fine with that. But what are you doing with this data and how are you leveraging it in a first class way. And we have otel we've throughout our analytics and our topology and other things. And also you touched on Turbonomic in a few of these, and I can elaborate on those later, but there's some other interesting acquisitions we've done too. And we're really driving these things together and bringing cost data in, for example, with KubeCost. So we've rounded out Instana and built around that core of best in Cloud Native and brought all of the what are now fundamentally expected capabilities around an observability solution.

Mirko Novakovic: Yeah, that sounds like a good idea. I know there were It's not easy, right? As a startup, you have to pick your bets and then you start. And observability is so big. And you just mentioned security, which is also interesting. I would like to have your opinion here how those categories are merging together. Right. If you look at Datadog, Dynatrace, the other competitors, Datadog for a long time was a modern monitoring, and now it says modern monitoring and security. And you see that more and more that it seems like those categories in parts are converging right into a platform. And as far as I understood, you are also going that path. Right. Having. Yeah some sort of checks on logs and other things.

[00:09:42] Chapter 5: Convergence of Observability and Security

Bill Lobig: Yes. So it's interesting, you know, Datadog did rebrand and they went hard at SIM, you know, security event and incident management. You know, it's funny, people talk about platforms, but I don't believe anyone buys a platform. Platforms emerge through the synergy of adjacent use cases, and it makes it easier to get to use case B from A than it did to originally get A. So if you have these things, and if you get into this SIM and security, the user persona of that is pretty different in my opinion, than, you know, someone who, for example, is looking at cost considerations in an SRE context. But they're all linked in that spectrum of, you know, DevSecOps. What we've focused on heavily is less on the sim use case, more on vulnerability management and shifting left. I hate this term, but it's used. It's a kind of a cliched term helping developers be more proactive in what they do. So if you think about, in my opinion and you're the expert because you founded this company, I, I inherited it largely. The APM market, in my opinion, was built around an assumption that something will go wrong and a human needs to root cause analysis, you know, trace it and fix it. Okay, well, how do we get some of those considerations into the pipelines and the ci CD processes so that we can have better a, b testing, release, rollback and things of that nature to catch these issues much earlier, or even allow real time debugging and troubleshooting of issues by developers in a production context, which is probably a whole other hour discussion there on the security considerations and such there. So we're looking at vulnerabilities, security, code quality, things of that nature on the, on the, you know, pushing towards the developer side of things.

[00:11:28] Chapter 6: Continuous Performance Improvement Challenges

Mirko Novakovic: Yeah, that makes sense though I my experience is that having performance in CI CD requires a really high maturity of the customer, because it essentially means that you need the tests, some sort of automated tests during that period, because if you otherwise it's more a static code analysis. But if you really want to do. I tried this a lot of times having this continuous performance improvement and but most customers are lagging behind those tests, right. They don't have them. And it's really hard also to keep them up and running. That's something I could see I work right because you do a code change and I could help you keep those tests up and running, right. Because I was a developer myself. That's the first thing. If you have high pressure and you need to get that fix out, you do your unit tests. But the automated test suite, you say, I can fix that later, right. And then and then once that starts, you will never get them up and running again. Right? That's kind of a big issue I see in the market.

Bill Lobig: I agree with you. I think the good news is people are motivated in that direction because what you're describing is test escapes, right. And you know, the pain comes back later in a different form. And this is where there's so many of these things converging, you know, code generation. We talked about Watson X and tools like developer assistance and like, how do we help them write those tests? Because one part is writing functionality, not tests, you know. And so all of this, I think, is part of driving automation through and improving application resilience throughout beyond observability tools.

[00:13:14] Chapter 7: Apptio Acquisition and IT Financial Management

Mirko Novakovic: Yeah. How do you see. That's something I'm just curious about. It's a cost perspective. You as Instana, you run on premise and in the cloud. So on premise, it's kind of you can move the problem of infrastructure somehow to the customer. Right? But in the cloud you can't. And one of the things I see as an issue, if you talk about AI and LLms, is if you use that on the amount of data that's produced in observability, it can create a lot of cost, because it's just I mean, you have probably billions of logs to process. And if you were to apply an LLM on each of those logs, or you have to do it really smart and really intentional, otherwise there's a high chance that you have an explosion of costs and the value compared to that. I don't know if it always matches. I don't know if you have seen that problem and how you address that.

Bill Lobig: I don't think it's hit mainstream yet to the point that it's affecting the majority. I think this idea of LLms and GPU and compute and cost associated with it in general. I mean, I'm looking at Nvidia's valuation. What is it, over 3 trillion now? I don't know anyone. I haven't met anyone who can clearly articulate the ROI. They were there. The promise is there, which is why all this investment is being made. But how are we really tracking those costs? And this is maybe a great opportunity to talk about you mentioned earlier bringing these things together in platform. You know, we acquired Apptio about 18 months ago, and I've been working for the leadership of Apptio, which is actually interesting. You know, IBM typically by company, they integrate them. I went into Apptio and it's all about, at the end of the day, getting to unit economics, which is every piece of cost, human infrastructure, licensed software, hardware. How does it all roll up into a business initiative? And then you can have accountability or responsibility all the way back through that, which is why we're bringing cost data into Instana, for example. So it's kind of like AI in powered by AI and then observability of AI. And you know, you had also asked, I'm kind of jumping back here, you know, what have we been up to in these last four years? GPU and LLM observability too. We've got some cool articles on, you know, in the Lang Trace community and some things of that nature as well to help people measure and understand these. But I agree with you. We can't just throw the hammer of AI at everything because there is big considerations to be had.

Mirko Novakovic: Talking about Apptio, I saw that acquisition. I know Apptio, but can you explain shortly to the audience what Apptio does and why it's important.

Bill Lobig: So Apptio is an IT financial management and FinOps company. And what is the difference in those two things. So there's kind of two dimensions. One is around traditional I'll call them CapEx CapEx and total cost. So if you think about Apptio Itfm, it's like a wide angle lens. It includes all of your costs from capitalization, depreciation, your most precious resource, your people, human labor costs, software, all of these things. So Apptio is the market leader in that they've coined and drive the TBM strategy. It's called technology business management. If you Google that, you'll see there's a an open source council around it. And then we also have the zoom lens if you will, which is within that is FinOps. And that is a discrete specific focus on cloud costs. And why is that different. It's because in the cloud. And it's probably obvious to everyone. Like the spend power, the delegation is very decentralized. Any developer on any day who has access to the cloud account can spin things up and down and drive massive cost considerations. And so that needs a new way of working. And you know, FinOps is a bit of a play on words, but it's kind of like DevOps. It's meant to bring developer and operations closer together. We still have development, we still have operations. But like they speak a common language and we've kind of improved that and getting finance and operations people to kind of be on the same page and driving that accountable spend is a really important part of this. And where it ties to observability is we are seeing those SREs, those DevOps, the people who live in these IT operations, these observability tools now want to see cost considerations natively in those tools because they're being increasingly asked to be mindful of that and to monitor and measure it on behalf of the financial obligations of the business.

[00:17:44] Chapter 8: Turbonomic and Resource Management

Mirko Novakovic: So essentially, Instana can feed data into Apptio to provide information about cost and infrastructure and metrics and usage and applications and all these things. Yes, that is.

Bill Lobig: A great synergy. I'm glad you brought it up. So Instana knows what all the application, dependencies and relationships are, and it can feed that to Apptio. And in this world of cost management, generally speaking, there's two ways to look at it. There's assumptive based allocation and consumptive assumptive means like assumption. You're making assumption right. So consumption is where you want to be because it's based on real usage. And I'll give you a great example. Kubernetes is highly virtualized infrastructure technology. If you have ten applications all running against the same cluster in pods or getting evicted and spun up and nodes are coming and going, how do you allocate those things accurately? Most companies just I'll take the number of applications, divide by that. And you know, everybody gets an equal portion Instana can tell you exactly how much workload and demand is coming from everything and then drive that more consumptive in the inverse Apptio can now feed cost data into the DevOps and SRE views of Instana. So it's just one more consideration there like performance, because if you zoom out, the reason there's waste in cloud is because people take an overly conservative posture to ensure application performance. Right? It's it's easier to get dinged for your budget being a little out of whack than it is. You know, the thing goes down during the critical period. And then you have a system outage and unhappy customer. So how do you balance performance and cost. Like that's what it's all about. And this is why I think observability at APM tools are so ripe for that integration.

Mirko Novakovic: Yeah. And I had an eye on this podcast, the CEO of Turbonomic and one of his claims, I remember also at IBM when he told that internally, because the obvious thing is to say Turbonomic drives with cost, right? But Ben is always saying, no, that's not the driver. It's performance. Right? Yes. You said people care more about their apps running smooth and with high performance then spending for it. Right. And he made the example. If spend would be the number one driver you can just shut down the application. Right. Then you have zero cost and. Yeah and I love that. But at the end of the day Turbonomic manages the resources and can. Then it could use Instana and Apptio information to then manage and balance those infrastructure. Correct.

Bill Lobig: Yeah, that's exactly it. And to your the comment you made I find amusing because I use that too. And I know, Ben, you know if it's primary job was to save costs, they would turn everything off and go home. Right. Like it's ridiculous to think about but that their job is performance. And how do you do it in a cost effective way? And you're absolutely right. Turbonomic is an application resource and performance management tool. It happens to size things down or up. And that has also been integrated, by the way, into Turbonomic in the fourth quarter is now a native integration with Instana. So we've brought that capacity based remediation technology there. So there's two ways in which that works. One is in an incident context, if I need to scale the thing up or down, I can now do that in a very integrated and elegant way with the AI of Turbonomic ensuring that you're going to have performance guarantee. But also if you think of not just Turbonomic, but any infrastructure as code, get any automated system that's changing things an APM or observability tool might see the CPUs just went in half and the CPU doubled and they're like, oh, an alert went off. Well, maybe that was intentional because I didn't need all those CPUs and I was wasting money. So now the APM tool observability can be smarter about these infrastructure changes and why they're occurring to minimize false positives.

[00:21:43] Chapter 9: Addressing On-Premise and Cloud Observability

Mirko Novakovic: Talking about Turbonomic as I mean, as far as I remember, it was initially basically built for managing VMware instances. Right. And, and I remember at the acquisition time it was very early around Kubernetes, but I can see that it probably evolved into that cloud native world. So it can manage Kubernetes now too, and understands the workloads and how is it integrated there?

Bill Lobig: Yeah. Turbonomic it's funny, I learned that the product was originally named VM turbo. That was like a little history tidbit there. But it's evolved greatly, much like Instana. So not only does Turbonomic now manage and optimize Kubernetes and does it in a way that is more powerful than horizontal pod scaling or vertical even in fact, the Red hat team is using it, right? And like, they're an OpenShift team. We've also invested heavily in optimizing what I'll call PaaS Hyperscaler or cloud service provider PaaS and beyond IaaS. So, you know, compute, memory storage, EKS, S3. We do all of that now. We see our clients demanding and we've implemented, you know, Aurora DB, redshift Databricks data. Right. Getting into optimization of of those SaaS native services that run on their so GPUs, even you know, these things are very expensive and are they being fully utilized. Your inference times latent or not. You know, all of these are areas where we've accelerated Turbonomic in the last few years.

Mirko Novakovic: And, and talking about VMs and servers and PaaS and on premise. Right. I think you are now almost the last vendor left who is supporting on premise. To be honest, I know that the newest Dynatrace version going cloud only. And I know a few of the developers still around. I'm still in contact with them, and I know you that you're heavily invested in making on premise super easy with Instana. And so how is your strategy there? I think I mean, you have probably a lot of governmental customers and others who have high security demand and they will want on premise. So how is your thought there and how do you make sure that this works smoothly with the SaaS infrastructure? I know how hard it is. That's why I'm asking the question, right? I always struggled with the on premise and SaaS keeping it in parallel, right?

[00:24:09] Chapter 10: AI Implementation Challenges and Opportunities

Bill Lobig: We still see a lot of interest in demand for, you know, we refer to it as self-hosted now, right? So you can run it, it's a container. You can run it where you want. And to the point on optimization or consumability improvements we've made. You know, a lot of clients say bring a container. Okay, I can run it locally, but if this is a monitoring tool, at the end of the day, it can't go down. So it needs to be highly available, which means I need Kubernetes. And not all customers is popular as Kubernetes is and as strategic as I see it, It's still early ish days in my opinion. I talked to some customer recently. I don't know who. I can't say their name anyway, and they said we have 1000 applications. Like 100 of them are on Kubernetes, right? There's a lot to be done. We continue to release feature parity SaaS and on prem at the same time, I know you mentioned Dynatrace there hanging in all the new stuff. Grail is going to SaaS. We're still doing it equally in both. It is hard. We have clients who want it. We've invested in making that ability to run that thing and self-contained way very easy using Linux, install embedded VMs so you don't need to stand up your own Kubernetes infrastructure. If you're not, you know, that mature yet in your journey along Kubernetes.

Mirko Novakovic: Yeah. That's cool. It's really cool. And how do you see AI in that context? That makes it even more complicated, right? If you are using AI, you have to provide that infrastructure also in an on premise environment.

Bill Lobig: So that we've not solved yet, admittedly. So for our AI use cases, we're inferencing from the cloud. Obviously that's a customer consideration or conversation. Do you want to send your thing there to train? Is it a rag pattern? Do you even need to send any data? You know, because some of these things like the known problem, known solution, we can get information and solve these, answer these questions, or give these recommendations without even using the customer's data because we've trained it on other sources. But that's not a solved problem at this point.

Mirko Novakovic: Yeah, it's same with Apple, right? They also want everything on the phone. But for a lot of use cases, they have to send the data to the Apple cloud, which they say is more secure. But it's something I can also see that it's a hard problem to solve. Right.

Bill Lobig: And I think it's interesting for Apple too, because now they're brought ChatGPT into Apple intelligence. So now they're even outsourcing some of the stuff over, you know, versus their privacy. And security has always been their primary thing. So it's interesting to see how that's evolving.

Mirko Novakovic: And how do you see AI and agents develop in the context of observability at the moment? My observation is that there is a lot of noise, but not so much real AI implementation with super high value yet. Right. You saw at the beginning you saw something like chatbots, right? Which could explain something. You saw natural language as a query language. But I haven't seen too many really groundbreaking use cases yet. How do you see that? What did you implement at IBM and with Instana and what do you see in the next? I think nobody can see further than a year at the moment. Right. It's hard to see how that develops. But what do you see in the next year or two?

Bill Lobig: I agree with you. This stuff is very high on that hype cycle, that hype curve. What do you call it? Peak of inflated expectations. But I do think, you know, as we get into that sort of plateau of productivity, at some point there's real value in there. So the areas we think that will emerge and that we're investing on. And by the way, one of the amazing things about IBM is, is our research team, like billions of dollars. These guys are just experimenting stuff. And we're looking at, you know, how can we use LLms for forecasting, anomaly detection, better multimodal telemetry analytics across log event. You know, melt metrics, log event traces? Agentic frameworks is kind of a hot new thing that, you know, we're incubating it. It's not there yet. But how do I have like a workflow of LLms where they're all specialized, right. Like we talk about if you go into the AI world, you large language models, small models, you mentioned the cost around these things. And IBM's very focused on these smaller, more targeted, fit for purpose models. And in the context of observability, what if I can have an, you know, an agent, if you will, that can do probable root cause and then it hands it off to the next one and says, well, for this problem, here are the potential actions or remediations. Okay, then hand it off to the. Now this model. Go do those actions and remediation. So another area now you mentioned its high degree of maturity. But maybe it's the Nirvana state or the Holy Grail. But getting to code repair is ultimately where it will become autonomic, you know, will it happen in our lifetime? I think so, but like anything, the people and culture piece is more important and harder than the technology piece. And I think that's what we're working through as an industry.

[00:29:08] Chapter 11: OpenTelemetry and Agent Use in Observability

Mirko Novakovic: I remember four years ago when I worked, still worked at IBM, I worked with Rama. She was on lab for AI and it was amazing what they were already doing at the time. If you look today in the lab environment. Right. And I was always like eager to productize that because they had really cool things. I remember, like pattern detection and stuff like that. Did you manage to transfer some of that AI lab stuff into the product because sometimes it's hard, right? Because as far as I remember, IBM lab is not built really to build products. It's really a scientific lab, right, where they can experiment and and they get not I don't know if it's still but as far as I understood, they're not incentivized to build products but really groundbreaking technology. Right. And so did you manage to get some of those things from Rama into the product.

Bill Lobig: Yeah. That's right. And that's by design right. Like if it's an engineering problem, which I define as just people and time like we know how to do it. We just need to do it. That's not the responsibility of research. And folks like Rama. They do prototype, they do experimentation, they prove ideas. We take those and turn those into commercial, off the shelf software. That's enterprise grade and scale. I would say eight out of those ten things don't make it, which is kind of the beauty of research, because not all of them are. You know, that's why it's research. The logging anomaly stuff you particularly mentioned on. Yes, that did make it. And actually that's part of the basis of how we've closed the gap where we didn't even have logging really four years ago, not in any any serious way. Now we do. And we haven't gone as far into like full log analytics or clickstream analytics like a, you know, Splunk might have. But for observability, AIOps infrastructure and even now the vulnerability use cases, you know, that has evolved into a really powerful set of foundational capabilities.

Mirko Novakovic: Yeah, I loved it. I saw the initial ideas and also saw it working in, in, in those environments. And it was amazing. Right. And I, I love to see that it made it into the product you mentioned. And I have to say, I totally agree with you. I always hated the word AIOps. Right? Because I never understood what it is. Because if you looked at the vendors who were either that were companies like Moogsoft at Bigpanda, which was basically event management, right?

Bill Lobig: Yes.

Mirko Novakovic: And so I was always struggling to understand, oh, the next big thing is AI ops and how did that work evolve? Because it was called Watson AI ops for years ago. Is it still there? How did the product evolve and how do you see that develop in the future?

Bill Lobig: The product is still named that. It's very hard to change those names. So, you know, because it takes another year for what happened to that thing. You know, the name changed, right? So that's always hard. It was coined after the industry usage of the term. And one of the things we learned since then. You're right. Very event incident management driven. You mentioned Moogsoft big Panda's out there. These have like manager of manager approaches. I will call them. I think if you don't if you don't own the data, you're at a disadvantage. Right. Coming in and saying now OpenTelemetry and these techniques will commoditize that out. But coming over the top as I got I. That's better than whatever the thing that collected your data has got. It I think is going to be hard. I feel strongly that AI is a feature of everything, as I mentioned earlier. So you in our strategy you will not see AI ops SaaS doesn't exist. We implemented that for self-hosted. There's a lot of use cases there. We have big business customers are enjoying that. It continues to innovate. But for SaaS, we decided to put all of this into Instana and just make Instana the AI powered observability solution, you know, for the future of those workloads.

Mirko Novakovic: Yeah, that makes sense. That's what's interesting.

Bill Lobig: I don't, I don't think it's true. It's like saying AI data. It's like AI ops AI applied to ops. I applied to what is AI applied to data mean? Like everything. So I'm not a fan of the term either.

Mirko Novakovic: And I have a question of OpenTelemetry just I mean, I'm a big fan of OpenTelemetry. Obviously we we at Dash0, we only do OpenTelemetry. But I have to say that in a lot of times I'm really missing the agent of Instana because it was so powerful. And OpenTelemetry is hard to implement, right? It's really hard to implement. It's not easy. It's not out of the box. It does not do a good job in auto discovery. So I'm not sure if you're allowed to tell the real numbers, but maybe in general, what would you say your customers, how do they leverage OpenTelemetry compared to the Instana auto discovery agent? Is that like OpenTelemetry is an addition to the agent? Do they replace the agent with OpenTelemetry or how? How do you see the adoption in your customer base, which are, I would say largely also really big enterprises, right? Are they really adopting it or are they still using heavily these auto discovery easy to use agents.

Bill Lobig: We're seeing it in addition to and since we added OpenTelemetry support a year ago, we've seen a massive, massive uptick in its usage. Like six x. And, you know, we could do the math on that. And you're right. It's... Was going to use the word crude. That's an unfair term. But there's more assembly required to your point. Right on implementing that. I embrace it heavily and welcome it because look, if you look at the industry and the market, you know, there's some £800 gorillas out there. Right. And they've largely locked in and expanded those use cases because of the stickiness of agents on these systems. And it's a switching cost. So if, if, if OpenTelemetry can supplant those and make it easier. And then the value goes to what you're doing with that information through your topology, your probable root cause, applying AI through correlation. That's a great advantage for us because, you know, we'd love to get some more of those £800 gorillas, in, in our remit. We also see a high demand for an OpenTelemetry based SDK, if you will, which is something we've introduced and using that to get back to some. And maybe this is part of that self-hosted use case we have like some of that older workload like the infrastructure monitoring, you know, so if Datadog infrastructure might but there's infrastructure that is kind of not CSP PaaS infrastructure. It's more traditional middleware kind of stuff. And how can OpenTelemetry be implemented there? And if the community's not doing it, how do you provide a connector so, you know, customers can do it themselves? So it's really an augmentation from what we see.

Mirko Novakovic: Is there an open telemetry SDK for COBOL?

Bill Lobig: That's a good question I don't know.

Mirko Novakovic: I don't know either. But that would be interesting. Right. Because I know the mainframe integration. And if you have like cakes or IMS.

Bill Lobig: Yeah, a lot of that's coming through Omegamon and these types of tools today. Yes.

Mirko Novakovic: Yeah, but not yet OpenTelemetry. But it could be, it would be cool. Right, to have a COBOL.

Bill Lobig: There's no reason why not.

[00:36:34] Chapter 12: Closing Remarks

Mirko Novakovic: Yeah. I mean, the protocol is standardized. Yeah, it should work. So, Bill, it was really fun hearing about Instana. Great job. It makes total sense to widen the platform. Synthetix rum logging I was added and the integration strategy. Right with Turbonomic with Apptio making it a. Yeah. You didn't like the word platform, but essentially it is a platform then, right?

Bill Lobig: That's what we're working towards. But it emerges from the synergies and we've got a lot of really great best in class technology acquired and organic coming out of research. Back to your Rama examples. And we're excited about unlocking new use cases and value for our customers. Perfect.

Mirko Novakovic: Thanks, Bill.

Bill Lobig: Thanks, Mirko. Take care.

Mirko Novakovic: Thanks for listening. I'm always sharing new insights and insider knowledge about observability on LinkedIn. You can follow me there for more. The podcast is produced by Dash0. We make observability easy for every developer.