 Hello everyone, glad to be here at KubeCon Europe 2024 in Observability Day and today I'd like to talk to you about the journey from observability to APM and taking cloud native up the stack. This is a tweet from KubeCon Europe last year. I'm Don Horvitz, I'm the principal developer advocate at logs.io, a cloud native observability platform that's based on popular open source cloud native stack. I'm also a CNCF ambassador and I also run the Open Observability Talks podcast. So if you're here, you're interested in the topic, so do check it out on your favorite podcast app or on YouTube. And you can find me everywhere at Horvitz as you can see here. And let's talk about APM. We tend to think that APM is old school, right? Observability is the new way to monitor microservices, cloud native, highly distributed systems. Funny enough, even the analysts are riding the wave. For example, you can see the Gardner Magic quadrant from APM shifting into APM and observability and then shifting, fading out APM altogether. But is it really old school? To be honest, until fairly recently, the cloud native observability stack has largely focused on rather rudimentary and siloed capabilities around logs, metrics and traces. We could individually analyze individual requests with tracing, then some text searching application logs, and then some custom metrics. But all of that is, you know, but a fraction of the well rounded view we've grown used to with APM back in the monolithic era. And you can see some examples of a far richer suite of capabilities in APM like real user monitoring, synthetic monitoring, incident integration, auto discovery, and much more. So on this lightning talk, let me touch upon how we can bring some of these APM capabilities into observability in general, and to the cloud native stack in particular, and also some wins that we've already achieved. And let's start with service performance monitoring. We need to have a basic red metrics on our microservices, request rate, error rate, latency. And I'm glad to say that we've done work at OpenTelemetry Collector to give the ability to calculate these metrics on the fly from spans from trace data. This is called the span metrics connector in OpenTelemetry. And as you can see here in the diagram, essentially you ingest the trace data in the tracing pipeline, and then via the connector, you roll up these into metrics that you emit as time series data from the collector. And once you have the span metrics available, then it can be queried and visualized on the back end. Going back to the cloud native stack, we have Yega already supporting these with a new monitor tab, as you can see here on the screenshot. And I'm proud to say that my folks at Logs.io contributed the original features of this SPM. So check out the QR code for the background and the work on Hotel and Yega in this part. Also, service mapping is an important element, and this also can be derived from the trace data. Here again, using Yega's example, you see how it generates the system architecture, essentially the dependency graph between the microservices based off of the span data. Real user monitoring is far behind still. There is a SIG, a special interest group to add real user monitoring to open telemetry. The work is ongoing on browser events, mobile events, semantic conventions and more. As you can see some of these, check out the Hotel client side telemetry Slack channel on the CNCF Slack, and also the QR code here for the charter of the SIG. Profiling is another important pillar, analyzing the resource utilization, CPU, memory, disk, whatnot, on different various parts of our code. Hotel is in the process of introducing profiling as a new signal. The tweet you can see here on the screen is from September last year, 2023, with the data model version 1 release, then there was a version 2 release back in November, so a lot of work is being done on this one. Check out the QR code and also the CNCF Slack channel for more on this. And there's much more obviously in the lightning talk, I can't cover more, but we can talk about more views like synthetic monitoring, SLO monitoring, business monitoring and so on. We also need to give thought about instrumentation. APM has a mature one and we're still not there with the observability stack. And then there's on the back end data analytics. This is where the vendors can play a key role and bring the differentiation on top of what is emitted from hotel and also AI, machine learning to get anomaly detection, noise reduction, assisted root cause analysis and so on. So in essence, this lightning talk is a call for us as a community to see how we take open observability up the stack. I'm Doton Horvitz, thank you very much for listening and may the open source be with you.