Yuval Lev’s Post

“eBPF-driven programs are the most exciting turf for platform engineering teams that are responsible for networking, security, and observability, and many who use these programs do not need to know anything about the underlying eBPF abstractions that make them possible.” (Link in comments.) It’s true–#eBPF 🐝 is a game-changer, enabling low-overhead data capture in complex systems without the hassle of manual setup. Lightweight, secure, and non-intrusive, eBPF lets you gain visibility into parts of your environment that would be otherwise difficult to instrument. But eBPF-based data collection is just the start. In order to make the data useful, you still have to make sense of it – by constructing a topology and configuring dashboards to define key relationships and business flows across your environment. This is often the step where eBPF – on its own – falls short. Data without context means nothing. That’s where Senser helps. We use eBPF technology for data collection, but we go further by automating the mapping of relationships across different layers of your environment–APIs, applications, network, and infrastructure. The result is a context-rich system map sitting on top of robust eBPF data collection. A proprietary in-house graph-based ML helps #SREs automatically identify and troubleshoot issues. Senser is laser-focused on minimizing manual efforts in root cause analysis, allowing your team to focus on solving problems and innovating. How are you managing data context in your observability practices? Share your thoughts!

It’s the Kubernetes curse of success. 👿 48% of Kubernetes users report “having problems choosing and validating which infrastructure components to use within their production environments” (h/t Spectro Cloud, link to highlights of the report in comments). In short: increasing K8s maturity and thriving ecosystem => complexity and headaches around tool choice and configuration (and tools playing nicely with each other). Another headache? Monitoring and troubleshooting your Kubernetes stack, especially with bigger deployments (for reference: 57% of users have 20+ cluster in productions in 2024, up from 35% in 2022 📈). My co-founder Amir Krayden wrote a nice piece on basic observability concepts and challenges with Kubernetes… well worth a read 👀 if you’re struggling with K8s observability complexity (links in comments).

“One problem with anomaly detection is that it is not business-aware, and, if not applied carefully, it may end up creating yet more alert fatigue.” Nice roundup on the state of AI in observability (link in comments). Key takeaways: - Root cause analysis – and reducing manual troubleshooting – are among the leading use cases for AI across the observability landscape. - Context is key. The context provided by tribal knowledge (“this microservice has this known dependency that’s created issues before”) and the connection to business objectives and outcomes (“we always experience surges in user behavior on Black Friday”) is hard to replicate via AI – but essential for efficient problem-solving. - The vision of an AI #SRE is still quite a ways off – most vendors in the space are still experimenting with incremental AI use cases rather than full intelligent automation. Observability folks out there – what other AI trends do you see in the space?

For those of us whose travel plans can’t wait… 🏖️ The Expedia outage (link in comments) is a prime example of why modern enterprises need AI-powered observability and root cause analysis. Initially, Expedia cited a "software issue" for the global disruption that impacted bookings, customer service, and operations for over 24 hours. But they later backtracked, unable to pinpoint the root cause across their distributed systems. This highlights a *major* challenge. As environments grow more complex with microservices, containers, serverless, etc., there are exponentially more potential failure points. Traditional monitoring tools struggle to connect the dots and quickly identify the root cause(s) behind major incidents. AI/ML models trained on telemetry data from the full-stack can automatically map service dependencies and use advanced analytics to trace the impact path. This allows teams to rapidly pinpoint the root cause(s), understand business impact, and prioritize mitigation steps – reducing costly downtime. The result? More 🍹, less 🤯.