Observability Summit North America 2026
2026-05-21..2026-05-22
Minneapolis, United States
as the organisers put it
Observability Summit bring the observability community together to collaborate and advance open source observability tools & practices.
| title | author | length |
|---|---|---|
| Keynote: Welcome + Opening Remarks | 03m | |
Welcome and housekeeping for Observability Summit 2026 in Minneapolis with sponsor thanks to Chronosphere, Palo Alto Company, Datadog, Open Search and diamond… | ||
| Sponsored Keynote: Zero-Code Observability: Close the Coverage Gaps That Cause Outa... | Eden Federman | 04m |
Claims a deployable eBPF-based instrumentation pipeline that emits OpenTelemetry traces with no code changes and under 1% overhead, promising cluster-wide… | ||
| Sponsored Keynote: The Work Before the Magic: Autoremediation Readiness | Alok Bhide | 08m |
Autoremediation succeeds only with all the data (metrics, logs, traces ingested via OpenTelemetry), structured context (normalized log fields, service maps… | ||
| Keynote: 10 Million Spans Per Second: Lessons From Scaling OpenTelemetry at Reddit | Trevor Riles | 26m |
Reddit moves 8–14 million spans per second with an OpenTelemetry pipeline through Kafka into ClickHouse and describes how sampling, batching, and ingestion… | ||
| Closing Remarks | 01m | |
Day one wraps with a recap: attendees report many new takeaways, extended networking and community building, and multiple panel insights credited to program… | ||
| Keynote: Welcome Back + Opening Remarks | 01m | |
Morning welcome opens day two of Observability Summit 2026 and notes the MC role with Datadog MC introducing a brief agenda. The remarks highlight concrete… | ||
| Sponsored Keynote: OpenSearch - See Everything: Open Observability for Agentic AI | Anirudha Jadhav | 06m |
OpenSearch releases a stitched observability stack that combines OpenSearch, OpenSearch Dashboards and Data Prepper with Prometheus integration, unified… | ||
| Sponsored Keynote: Datadog - Every Byte Counts: How Protocol Design Shapes the Cost... | Amanda Sopkin | 07m |
Telemetry volumes grow about 35% year over year for the average enterprise, driving ingestion, storage and especially egress costs. A new protocol, OTAP… | ||
| Keynote: Tracing the Agent's Mind: Extending OpenTelemetry for Deep... | Mustafa Dayıoğlu, Zeyno Dodd | 27m |
Production agents make risky tool calls; this work introduces the gen-ai-otel extension that adds a gen_ai.agent. attribute namespace capturing tool-selection… | ||
| Closing Remarks | 01m | |
Two-day Observability Summit in Minneapolis concludes with thanks to speakers, program committee, sponsors, event staff, and attendees and notes applause and a… | ||
| Let Them Eat Bugs: Practical Showcase of Agentic Issue Resolution | May Walter | 26m |
Hud builds a runtime intelligence layer that detects production errors, queries ClickHouse for function-level context, runs opinionated investigation skills… | ||
| The Legend of Config: Breath of the Cluster | Henrik Rexed | 23m |
Demonstrates that OpenTelemetry plus eBPF (OBI) reveals network flows but misses route and policy context, so a custom Gateway API processor is built for the… | ||
| Breaking Free from Vendor Lock-In: Nubank DIY Observability Success | Diego Rocha, Otavio Valadares | 27m |
Nubank builds an in-house logging pipeline that ingests 1.5 petabytes and 1.5 trillion log lines per day, peaks at 45 GB/s, and reduces vendor costs by ~50%… | ||
| eBPF Application Instrumentation for Java: Challenges, Design, and Real... | Endre Sara, Stephen Lang | 24m |
eBPF daemonset inspects TCP at the kernel with kprobes and uprobe mix and injects a tiny Java agent into running JVMs to extract connection metadata and… | ||
| Lightning Talk: A Drop-in System To Accelerate Metrics Observability by 100x Us... | Milind Srivastava | 12m |
Claims 100x reduction in query cost and latency by adding a drop-in proxy that implements the Prometheus HTTP API and ingests raw metrics via remote write… | ||
| Lightning Talk: GPU-Scanner: Extending CNCF Observability for Multi-GPU AI Workloads | Ritika Gupta | 09m |
GPU-Scanner is an open-source Kubernetes extension that adds active health checks (PyTorch benchmarks for baselining) and passive health checks (30–40s probes… | ||
| Applying Observability to the Internet of Living Things (IoLT) | Sophia Solomon | 20m |
A virtual cat called Meowy is built in Go, instrumented with OpenTelemetry and exported via OTLP to an Elastic/ELK backend and Kibana, with a batch span… | ||
| Observing the Observers: Bringing OpenTelemetry to Autonomous AI Agents | Abdel Fane | 23m |
Telemetry is blind to agent identity and authorization so attackers exploit MCP servers: 206,000 attack events recorded in honeypots with 75% targeted MCPs and… | ||
| Devs, Transform (Your Data) and Roll Out!: Learning and Leveraging OTTL | Reese Lee | 16m |
OpenTelemetry Transformation Language (OTTL) runs in the Collector’s processing stage and powers the transform, filter and tail sampling processors, letting… | ||
| How Observability-First Development Lets You Ship Agents in Weeks... | Anirudha Jadhav, Kevin Fallis | 17m |
Claims that building observability in from day one cuts agent time-to-production from 12 months (retrofitted) to 6 weeks (observability-first) and shows the… | ||
| Implementation of Unified Observability at Scale From Scratch | Ahmed J. | 23m |
Built a unified observability platform across 200+ sites handling 2,000 VMs, 10,000 network devices and 300 firewalls using Grafana Alloy in clustering mode… | ||
| One Size Does Not Fit All: A Polystore Architecture for Logs and Traces | Suman Karumuri | 25m |
Claims a polystore routing logs and traces to specialized engines (Lucene/OpenSearch, DuckDB) via Kafka and S3, with a unified OpenSearch-compatible API and… | ||
| Beyond Dashboards: Architecting AI Agents for Autonomous Observability | Divya Mahajan, Achin Gupta | 28m |
Claims an RCA pipeline that reduces typical 90 minute investigations to ~30–90 seconds by combining deterministic math with LLMs and RAG over a vector DB… | ||
| Lightning Talk: Observability Debt: When Telemetry Stops Telling the Truth | Spoorthi Palakshaiah | 10m |
Claims that observability debt causes telemetry meaning to drift so dashboards and alerts lie: a simple API + queue + async workers model shows P99 latency… | ||
| Lightning Talk: Show Me the Money: Metrics Edition | Brian Davis | 10m |
Shows a concrete method: add Prometheus metrics with labels for component, data type and tenant ID, scrape daily, and join metric totals to the cloud cost and… | ||
| Show Me the Receipts: A Forensic Hunt for Observability | Mostafa Radwan | 19m |
Shows a repeatable forensic audit that traces resolved incidents back to the exact telemetry that enabled resolution, yielding 30% log reduction for one… | ||
| Exploring Observability with MCP Servers | Tiffany Jernigan | 26m |
Open-source Grafana MCP server exposes Grafana, Pyroscope and connected backends to LLMs so assistants can list services, query failing traces, add flame… | ||
| Why Are Your AI’s Decisions Hard To Explain: Trace Every Decis... | Dhiraj Kumar Jain, Vikash Agrawal | 27m |
Argues that observability must trace every agent decision, proposing one span per decision recorded via OpenTelemetry and ingested into OpenSearch + Data… | ||
| From Data Dumps To Smart Context: Building MCP Servers That AI Can Actually Use | Thomas Johnson | 20m |
Most MCP servers dump every metric; this one instead captures full‑stack traces, logs, requests and headers and pre-correlates sessions so agents get a… | ||
| One Pane to Rule Them All: Uniting the Prometheus Community with O... | Anirudha Jadhav, Kevin Fallis | 24m |
OpenSearch now ingests Prometheus metrics natively and surfaces them alongside logs and traces in OpenSearch Dashboards using Data Prepper and OpenTelemetry… | ||
| Whats the Best Way To Reduce Storage Requirements Without Losing Insights? Push AI To... | Alex Degitz | 25m |
ElastiFlow describes an OTel native edge processing node that runs an agentic AI (often gpt-oss-20b) and stores signals in local DuckDB storage, making routing… | ||
| Policy as Code Meets OpenTelemetry: The Next Frontier of Observability | Christopher Voisey | 25m |
Combines Open Policy Agent and OpenTelemetry to treat authorization outcomes as structured decision telemetry, emitting a privacy‑preserving fingerprint or… | ||
| Unified End-to-End Observability: How Comcast Generates SpanMetrics at Enterp... | Raghu Vamshi Challa | 26m |
Comcast describes migrating 350 applications from AppDynamics to an OpenTelemetry stack that generates RED metrics by converting spans with the OpenTelemetry… | ||
| Lightning Talk: From Collector To Terminal: A Better Way To See Your OpenTelemetry Logs | Jon Reeve | 09m |
Gonzo connects an OTLP-capable app directly to a terminal UI so logs stream in with structure, attributes and heatmaps without a Collector or JSON munging… | ||
| Lightning Talk: Summarizing the Noise: LLM Observability With Open Data Hub, VLLM... | Twinkll Sisodia | 11m |
An open-source AI observability summarizer built on Open Data Hub deploys LLMs via vLLM and KServe, scrapes metrics from Prometheus/Thanos, traces from Tempo… | ||
| AI-Powered Root Cause Analysis at Scale: From Theory To Production... | Letícia Mota, Yevgeny Gladun | 27m |
Nubank builds a Python orchestrator that runs in Kubernetes, using ChatGPT-4 then GPT-5, Prometheus, Honeycomb, GitHub and a Blueprint API to do autonomous RCA… | ||
| OpenTelemetry GenAI in Practice: What the Spec Says Vs. What You Actually See | Zach Groves | 24m |
OpenTelemetry GenAI semantic conventions (notably v1.37 and follow-ups) are usable but adoption is uneven: some implementations emit clean 1.37+ spans (OTEL… | ||
| Secure by Design: Rethinking Test Credentials for Synthetic Monitoring | Katie Kodes | 23m |
A near-miss leak exposes full bank and active session cookies from automated browser tests and argues that commercial observability platforms now run Selenium… | ||
| The Full Picture: Visualizing Service "Fullness" To Rethink Saturation Prevention | Tal Nordan | 27m |
Proposes exposing each instance’s live concurrency limit and its current in‑flight requests by extending Envoy and leveraging eBPF and Cilium to read… | ||
| When the Cloud Fails: Debugging the "Undocumented" | Dhruv Jain | 21m |
Migration to a new cloud project triggers a production-only failure that cuts message delivery by about 7% drop for merchants and ~4% for drivers; targeted… | ||
| The Speed of Metrics, the Fidelity of Traces: Architecting Post-Collection Aggregation | Zack Owens | 26m |
Claims that post-collection aggregation gives metric-like query speed over long windows while keeping raw traces intact, demonstrated with ClickHouse… | ||
| Taming Tenancy, Cost and Architecture at Collibra Through OpenTelemetry and Our... | Alex Van Boxel | 24m |
Costs skyrocket when giving each tenant its own container, so Collibra switches to multi-tenancy and saves money while keeping per-tenant attribution by… | ||
| Panel: Telemetry That Matters | Virginia Diana Todea, Antonio Jimenez Martinez, Laura Luttmer | 30m |
Claims that >50% of metrics are never queried and one customer hit $2M in telemetry overages motivate concrete fixes: use OpenTelemetry collectors and the… | ||
| Lightning Talk: Beyond Billions: Operating Thanos, Prometheus & OpenTelemetry... | Narendra Reddy Sanikommu | 11m |
Operating a metrics stack at 3.5 trillion samples/day (≈40M samples/sec, 2B active series) exposes failures not seen at smaller scale: Thanos Receivers return… | ||
| Quantiles at Scale: Choosing the Right Estimation Algorithms for Observability | Mike Shi | 25m |
Benchmarking on 7 billion rows in ClickHouse shows exact sort costs 39 GB RAM and 15 seconds and is impractical; Prometheus native histogram schema 4 yields… | ||
| Taming Observability at Scale in a Multi-Cluster Kubernetes Platform at Bloom... | Joe Nathan Abellard | 24m |
Bloomberg builds a managed multi-region Karmada control plane stretched across multiple host clusters for HA and disaster recovery, using per-cluster… | ||
| The Invisible Tax: How Data Format Conversions Drive up Telemetry... | Cijo Thomas, Joshua MacDonald | 22m |
Measurements show OpenTelemetry SDKs (Go and Rust) spend most CPU on format hops — 80% spent on SDK conversion in Rust versus less in Go — and collectors add… | ||
| Don't Let Users Find Your Outages: Synthetic Monitoring for Kubernetes Platforms | Kate Agnew | 23m |
Kuberhealthy runs recurring check pods every five minutes that create resources (Argo CD deploys a demo app, ingress + Route 53 record, then the check queries… | ||
| Inside the Telemetry Data Plane: Constraints, Tradeoffs, and Scale | José Lecaros | 29m |
Telemetry data planes confront three hard limits: memory, scheduling, and the throughput/latency budget. Fluent Bit-style agents use 2 MB soft chunk limits and… | ||
| nothing matches | ||