SREcon26 Americas
2026-03-24..2026-03-26
Seattle, United States
who it's for
site reliability engineers, SRE managers, platform engineers, incident response leads, observability engineers, cloud engineers, security engineers
topics
chaos engineering in production; developer experience for incidents; cloud right-sizing and overcommit; RocksDB SST offline serving; AI agents for log analysis; KubeVirt edge modernization; zero trust internal access; compliance as code with OPA; kubernetes conntrack troubleshooting; observability for LLMs; postgres high availability architecture; intelligent load balancing in Kubernetes
| title | author | length |
|---|---|---|
| Taming the Unpredictable: Reliability in Chaos | Michelle Brush | 43m |
Claim: AI-generated code will explode system size and unpredictability, forcing a shift from point-in-time reviews to continuous experiments and risk-first… | ||
| Mean Time to WTF: Why Developer Experience Frameworks Belong in Your Incident... | Nicole Forsgren | 41m |
Argues that applying SPACE and DORA to SRE work reveals measurable operational friction and that a new metric, Mean Time to WTF, (time from alert to “I… | ||
| Infinity Is Not a Strategy: Right‑Sizing the Cloud | Praval Panwar | 32m |
Argues the “cloud is infinite” myth and gives concrete replacements: use ARIMA forecasts and LSTM-style continuous load models, pick overcommit K from… | ||
| Low Latency Serving of Offline Data: Efficient, Safe, and Reliable Data... | William Schor | 32m |
Preprocesses offline Iceberg data into RocksDB SST files, stage them in S3, then load those files onto dedicated serving nodes with lease-based partitioning… | ||
| Autonomous Policy Validation: Building AI Agents to Analyze Logs and Identify... | Paul Huang | 36m |
Built an AI agent pipeline that converts logs into a traversable graph working memory, using hybrid parsing (deterministic regex/schema + LLM) for entity… | ||
| From ESXi to Kubernetes: Modernizing 1,400 Edge Locations with Open Source | Jasjit Singh, Kanika Nayyar | 15m |
1,400 retail edge sites migrate from ESXi to Kubernetes without buying new hardware by repurposing an existing node, rolling out a KubeVirt bridge to run… | ||
| Beyond Blanket Freezes: Enabling Safe Innovation During Critical Events at... | Prachi Jain, Sandhya Narayan | 17m |
Netflix replaces blunt deployment freezes with a data-driven, automated bypass wired into CI/CD that scores each change by tiers 0–3, four risk signals… | ||
| The Zero Trust Odyssey: Our Journey to Modernize Internal Access | Nathan Handler, Pratik Lotia | 37m |
Replaces legacy NGINX proxies with Cloudflare (WARP, Argo tunnels and Access), using Route 53 partial CNAMEs and ExternalDNS to avoid replatforming; automates… | ||
| Shift-Left Compliance | Bharath Modhipalli, Karanveer Anand | 32m |
Shift-Left Compliance makes regulatory requirements an intrinsic system property by turning rules into machine-readable policies and enforcing them in CI/CD… | ||
| Executing Chaos Engineering in Production at a Critical Financial Institution | Luiz Siqueira, Leonardo Siqueira Marques | 27m |
A financial cloud platform running Kubernetes, Redis, Kafka and external APIs adopts a reproducible chaos platform with governance guardrails, safety windows… | ||
| From Thundering Herd to Zero Outages: Building Reliable Inventory Sync | Rushikesh Ghatpande | 27m |
A production inventory sync protocol that uses a 5-stage FSM (connection, app-init, sync-handshake, full sync, delta sync) runs on a tiered topology with a… | ||
| The Critical Resource Is You: Practical Destressing for On-Call Engineers | Beth Adele Long | 15m |
Claim: stress lives in the body and engineers can use simple, measurable techniques to shift autonomic balance. Four practical tools are offered: body scan to… | ||
| AI Agents for Incident Investigation: The Good, The Bad, and The Ugly | Vladyslav Budichenko | 16m |
AI agents connect to observability stacks and coding engines to compress investigation time and sometimes cut MTTR, with about 50% of teams running pilots and… | ||
| Three Lies We Tell Ourselves about Disaster Recovery and What to Do about Them | Colette Alexander | 19m |
Claims three concrete failures: fantasy DR plans rarely work, operators face ambiguity, and DR’s main value is learning not perfection. Examples include… | ||
| Escaping Version Skew: Formalizing Compatibility in a World of Partial Rollouts | Robbie Ostrow | 32m |
A concrete method to avoid rollout breakage: detect breaking schema changes automatically, generate strict writer types and union reader types, and gate merges… | ||
| The Case of the Misnamed Cities: CAST Analysis of a Google Maps Incident | Ruben Barroso | 35m |
Google Maps displays nine misnamed cities after an imported US Census file uses “balance” names, manual fixes and an automated rollback conflict, and the… | ||
| Beyond Loss and Accuracy: Closing the Observability Gaps in AI Training with... | Yuxuan Jiang, Ryan Huang | 33m |
TrainCheck is an open-source Python module that instruments PyTorch training via monkey-patching, records API traces and parameter hashes, and checks inferred… | ||
| Stop Reading Changelogs: Safer Kubernetes Upgrades with Simulation | David R. Morrison | 35m |
An open-source tool called SimKube records production activity as a timestamped YAML trace and replays it into a simulated cluster built on KWOK (Kubernetes… | ||
| From Chaos to Confidence: How SREs Can Leverage 50 (and Counting) Failure... | Rohan R. Arora | 27m |
Built a sandboxed Kubernetes evaluation called IT Bench that replays 50+ production-inspired failure scenarios using Helm, Ansible, an open observability stack… | ||
| When the Cure Is Worse than the Disease: Metastability in Recovery | Todd Porter, Aleksey Charapko | 37m |
A global message bus called scribe (Kafka-like) experiences a cascade where issuing thousands of hourly backfills to fix missing partitions actually amplifies… | ||
| Resilient Observability at the Retail Edge: A Lightweight, Scalable, and... | Prakash Velusamy | 41m |
Framework uses K3s plus OpenTelemetry collectors, Fluent Bit/Fluentd, Tempo and Grafana to replace Prometheus at the retail edge, narrowing to 25 core metrics… | ||
| Ghosts in the Interview Loop and Avoiding AI Taylorism | Andrew Hatch | 34m |
ThousandEyes encounters in 2025: 300 applicants for a midlevel SRE, coding rounds produce AI-generated perfect code that matches ChatGPT verbatim, so the… | ||
| So You Want a New Incident Commander—Lessons from Building Incident Response... | Vanessa Huerta Granda | 36m |
Organization reframes Incident Command as sociotechnical leadership and builds a small, dedicated IC team trained with shadowing and reverse-shadowing… | ||
| Epistemology of Incidents and Problem Solving | Jack Kingsman | 35m |
Start-to-finish playbook for high-pressure incident work with concrete rules: codify alert routing, severity matrices and SLO intent; use Google SRE incident… | ||
| Human Factors in the Age of AI Ops: Re-Engineering Trust between Humans and... | Eddie Redick | 33m |
Teams are deploying AI faster than they build trust: 68% of orgs have AI widely, yet only 16% fully trust it for decisions and 73% of alerts are false… | ||
| The Unconspicuous Role of Conntrack in Kubernetes Networking | Ricard Bejarano | 30m |
On Sept 4, 2025 a 100x DNS request spike and Istio’s aggressive service entries refresh (every 30s) combined with Envoy reusing one socket causes all pods to… | ||
| It's Not Always the Network (But Here's How to Prove It): Kubernetes Packet... | Mitsuhiro Shibuya | 34m |
Practical playbook for proving whether networking is at fault: three concrete capture strategies and when to use each. Use tcpdump on eth0 from the node to… | ||
| Precision Over Proliferation: SRE Approach for Leaner, Smarter and Data-Driven.. | Md Shaghil | 31m |
Claims a 40%+ reduction in metrics cost by building a metrics management framework: a Dollar per Query efficiency metric and usage dashboards, a quota system… | ||
| Observability for LLMs: Understanding What’s Happening Under the Hood | Salman Munaf | 23m |
LLM systems require tracing the whole workflow from prompt construction through retrieval, tool calls, model invocations and safety checks, not just 200s and… | ||
| Lightning Talks | Bala Subrahmanyam Kambala, Charit Upadhyay, Apparao Boddeti, Philip Rowlands, Emmanuel I. Obi, Khushboo Nigam | 26m |
Six four‑minute reports cover concrete failures and fixes: a root CA roll reveals that root cert rotation is a distribution problem — services reload trust at… | ||
| The Gashlycrumb Tinies of AI Networking You Must Know (or Languish!) | Lerna Ekmekcioglu | 30m |
Demos show an NCCL all‑reduce hitting 1.47 Tbps on 16×100 Gb NICs until a second cross‑rack job cuts per‑NIC throughput from ~92 Gb to 50–30 Gb, driven by… | ||
| How We Built Protosockets to Go Beyond HTTP and gRPC | Pratik Agarwal | 24m |
Built a lightweight, message‑oriented layer over raw TCP called Protosocket (Rust, Tokio) that drops HTTP/2 and gRPC overhead, uses per‑connection buffers and… | ||
| Operationalizing Key Management for Regulatory Compliance and Emergency Response | Swetha Srinivasan | 26m |
Describes a concrete, repeatable system: a state-driven automated pipeline that adds a new key, pushes the new public key to all verifiers, waits for… | ||
| How Security Incidents Are Different ... and How They're Exactly the Same | Laura de Vesine, Alec Randazzo | 39m |
Scenario shows an SSO phishing proxy that yields AWS access, attacker downloads S3 data and demands crypto, detection often misses it until a ransom letter… | ||
| Unlock High-Frequency Deployments without Blowing Up Prometheus | Ganesh Vernekar | 16m |
Prometheus gets OOMs from Kubernetes rollouts because label changes and pod churn create thousands of stale time series that stay in memory until periodic… | ||
| How We Debug 1000s of Databases with AI: Lessons from an AI-Assisted Database... | Annie Zhou, Sophie Zhang | 17m |
An AI-assisted platform debugs thousands of MySQL and TiDB instances across 70+ regions and three clouds by consolidating metrics, logs and KB articles… | ||
| Building SRE Culture (without SREs, Technically) | Chris Harding | 15m |
Shows a real failure: a red sign build 37.2 gets rolled out accidentally, the platform lacks rollback so players download 53 GB downloads twice, many lose… | ||
| Reliability Engineering for Hybrid Robot-Cloud Systems | Jeff Corpuz, Rian Bogle | 14m |
Arc fleet management ties a cloud control plane to roaming robot nodes, handling intermittent MQTT sessions with QoS, defining robot SLOs and shaving MTTR via… | ||
| 5 Wrong Hypotheses about PostgreSQL Multi-Transaction Locks | Clint Byrum | 31m |
An index change triggers a vacuum ANALYZE and then the DB hits pathological SLRU aggregate accesses growth (millions/sec), locking up queries; killing long… | ||
| Designing Layered High‑Availability Architectures for PostgreSQL on a Budget | Umair Shahid | 36m |
Claims a practical framework that composes incremental Postgres HA layers—offsite backups, WAL shipping, hot async standby, sync replica and a cross‑region… | ||
| Intelligent Load Balancing in Kubernetes | Gaurav Nanda, Vincent Cheng | 35m |
Databricks replaces kube-proxy/DNS data-path balancing with a lightweight, push-based control plane that distributes EndpointSlice updates via xDS push-based… | ||
| Reliable OpenTelemetry at Scale: No Queue, No Problem | Tommy Li, Vlad Seliverstov | 34m |
Claims a production OpenTelemetry pipeline that avoids central queues, handling ~15 GB/s peak and trillions of events per day by keeping a fast direct path… | ||
| Keeping a Hypervisor Fleet Up to Date with Minimal Customer Disruption | Atalay Kutlay | 15m |
A production scheduling service models host reboots and VM migrations as interval variables and uses OR-Tools CP-SAT to optimize who and when to update, aiming… | ||
| The Evolution of Postgres Performance | Ben Dicken | 17m |
Postgres 18 adds an io_method (sync, worker, io_uring) option but PlanetScale benchmarks show io_uring is often slower than the worker thread mode and… | ||
| Talking to/with Machines | Audrey Simonne | 14m |
Engineers must treat LLMs as a separate audience and build explicit, shared assumptions as part of prompts and documentation rather than relying on vague… | ||
| Reliability Equilibrium: The Hidden Playbook behind SRE Influence | Daria Barteneva | 43m |
Frames SRE problems as game-theory failures and shows concrete fixes: use error budgets, canary deployments, progressive rollouts and reversible releases to… | ||
| The Power of Stories | Lorin Hochstein | 35m |
Argues that incident stories beat bullets for getting the right information into people’s heads, because good incident stories are anomalous and immutable… | ||
| The Ironies of AI² | J. Paul Reed | 41m |
Automation erodes manual skills, can hide degrading states, and deployed AI agents sometimes act without needed checks: an agent-to-agent localization change… | ||
| nothing matches | ||