SREcon18 Americas
2018-03-27..2018-03-29
Santa Clara, United States
as the organisers put it
SREcon18 Americas will take place March 27–29, 2018, at the Hyatt Regency Santa Clara, in Santa Clara, CA, USA. SREcon18 Americas is a gathering of engineers who care deeply about engineering resilience, reliability, and performance into complex distributed systems, and the scalability of products, services, and infrastructure within their organizations.
who it's for
site reliability engineers, engineering managers, incident responders, platform engineers, security engineers, devops engineers, network engineers
topics
DNS resilience and scaling; stable health checking algorithms; service level objectives and indicators; multi-region disaster recovery; multi-cloud migrations; Kubernetes platform as a service; distributed tracing lessons; capacity prediction methods; containerization operational war stories; database migration to MyRocks; security telemetry on Kafka; blameless postmortem practices
| title | author | length |
|---|---|---|
| If You Don’t Know Where You’re Going, It Doesn’t Matter How Fast You Get There | Nicole Forsgren, Jez Humble | 34m |
DORA’s four validated metrics are the actual claim: deployment frequency, lead time for changes, mean time to restore, and change-fail rate, measured across… | ||
| Security and SRE: Natural Force Multipliers | Cory Scott | 33m |
LinkedIn shows that embedding security under SRE yields concrete outcomes: security telemetry runs on Kafka data infrastructure for always-on visibility, an… | ||
| What It Really Means to Be an Effective Engineer | Edmond Lau | 28m |
Reports findings from interviews at Google Facebook Dropbox Airbnb Stripe Instagram and more and a 2015 book that reframes effectiveness as maximizing leverage… | ||
| The Day the DNS Died | Jeremy Blosser | 34m |
SparkPost runs 8,000 DNS queries per second and hits undocumented cloud limits that throttle at roughly 1,024 qps per network interface, which causes 90%+… | ||
| Stable and Accurate Health-Checking of Horizontally-Scaled Services | Lorenzo Saino | 34m |
Fastly builds a distributed health checker that classifies an instance unhealthy only when its behavior deviates significantly from the cluster, avoiding… | ||
| Beyond Burnout: Mental Health and Neurodiversity in Engineering | James Meickle | 40m |
Claims that mental disorders affect 25% now, 50% over a lifetime and that exclusion often happens in hiring not performance. Recommends concrete fixes: run… | ||
| Bootstrapping an SRE Team: | Aaron Wieczorek | 37m |
Combines two separate DevOps teams for Vets.gov and Caseflow into one shared SRE org that standardizes tooling (Prometheus and Grafana, Sentry, Travis/Jenkins… | ||
| Don’t Ever Change! Are Immutable Deployments Really Simpler, Faster, and Safer? | Rob Hirschfeld | 38m |
Shows an implementable path from configure-in-place to build-then-deploy by using image-deploy and in-memory netboot (demo uses Digital Rebar and CentOS)… | ||
| Lessons Learned from Our Main Database Migrations | Yoshinori Matsunobu | 39m |
Facebook migrates UDB from compressed InnoDB to MyRocks and cuts storage and server count roughly in half after a nearly yearlong rollout. Migration uses a… | ||
| Leveraging Multiple Regions to Improve Site Reliability: | Andrew Duch | 18m |
Jet moves from one Azure region to multiple by replicating a single event store replication as the source of truth, projecting caches and indexes locally, and… | ||
| Building Successful SRE in Large Enterprises—One Year Later | Dave Rensin | 40m |
Google’s CRE program claims enterprises are the easiest early adopters and reports concrete wins: Home Depot grows from 0 to ~800 services with SLOs in a year… | ||
| Working with Third Parties Shouldn't Suck | Jonathan Mercereau | 33m |
Argues third parties must be treated as part of the stack, not optional extras, and gives concrete practices: run RFI/RFPs, demand timely, direct answers… | ||
| When to NOT Set SLOs: Lots of Strangers Are Running My Software! | Marie Cosgrove-Davies | 14m |
Argues against shipping fixed SLOs to customers and instead provides built-in blackbox SLIs (deploy-test and app-availability), guidance, and a CRE-style… | ||
| Lessons Learned from Five Years of Multi-Cloud | Arup Chakrabarti | 34m |
PagerDuty runs the same services across AWS and Azure for about five years to hit a hard 50 ms latency requirement, deploying ~50 backend services across 10… | ||
| Help Protect Your Data Centers with Safety Constraints | Christina Schulman, Etienne Perot | 20m |
Covers concrete failures (a single rack decommission command that erased the CDN and a script that decommissioned every tunneling load balancer), the… | ||
| Real World SLOs and SLIs: A Deep Dive | Matthew Flaming, Elisa Binette | 20m |
Walks through concrete SLI/SLO recipes used at New Relic: measure data ingest by counting 200 responses and message-bus latency by comparing ingest timestamps… | ||
| How SREs Found More than $100 Million Using Failed Customer Interactions | Wes Hummel | 31m |
PayPal shifts from inward payment metrics to tracking Failed Customer Interactions (FCIs) by tagging requests with a correlation ID and surfacing HTTP 500s in… | ||
| Learning at Scale Is Hard! Outage Pattern Analysis and Dirty Data | Tanner Lund | 20m |
Microsoft Azure teams build a pipeline that starts with post-mortem text clustering (cleaning HTML with BeautifulSoup), fits ML models but finds model fitting… | ||
| How Not to Go Boom: Lessons for SREs from Oil Refineries | Emil Stolarsky | 17m |
Compares refinery controls to SRE practice with concrete techniques: pressure relief valves as spring‑loaded load shedding, isolation valves as bulkheading for… | ||
| Containerization War Stories | Ruth Grace Wong, Rodrigo Menezes | 26m |
Pinterest migrates a quarter of infrastructure to Docker over about a year, moving from brittle masterful Puppet to container images and a sidecar model. Key… | ||
| Resolving Outages Faster with Better Debugging Strategies | Liz Fong Jones, Adam Mckaig | 39m |
Claims that replacing static dashboards with on-the-fly query editing and trace-linked samples cuts mean time to resolve: use a query editor to layer peel from… | ||
| Monitoring DNS with Open-Source Solutions | Felipe Espinoza, Javier Bustos | 17m |
NIC Chile manages ~3,000 DNS qps per server and builds a real‑time monitor by integrating open‑source pieces rather than custom Hadoop stacks. Capture uses a… | ||
| Antics, Drift, and Chaos | Lorin Hochstein | 34m |
Claim that complex distributed services inevitably drift into failure is illustrated with concrete incidents: a missing functional-test annotation causes a… | ||
| Security as a Service | Wojciech Wojtyniak | 30m |
Facebook Production Engineering Security starts as four people and scales across two continents while supporting a 12-deploys-per-day production model. The… | ||
| Breaking in a New Job as an SRE | Amy Tobey | 20m |
Quick, practical checklist: negotiate relocation, RSU volume, laptop specs and phone reimbursement up front; generate a new SSH key and encrypt it; pair with… | ||
| "Capacity Prediction" instead of "Capacity Planning": | Rick Boone | 21m |
Uber models each service’s CPU as a function of business ingresses (completed trips, drivers online) using Spearman correlation to pick the strongest driver… | ||
| Distributed Tracing, Lessons Learned | Gina Maini | 34m |
Company builds a custom tracing platform in F# to track long asynchronous flows through Kafka and Event Sourcing, emitting compact telemetry events (trace ID… | ||
| Junior Engineers Are Features, Not Bugs | Kate Taggart | 22m |
Team experiments show hiring juniors forces fixes: at one org a team handling 85% of web and API traffic runs with one third bootcamp grads and builds a shared… | ||
| Approaching the Unacceptable Workload Boundary | Baron Schwartz | 38m |
Systems degrade across a fuzzy unacceptable workload boundary rather than a sharp limit; load is usefully defined as the sum of residence times (average… | ||
| Building Shopify's PaaS on Kubernetes | Karan Thukral | 33m |
Shopify builds a PaaS on Kubernetes running on Google Container Engine to consolidate wildly different runtimes, moving from ~20 services in 2011 to >400… | ||
| Know Thy Enemy: How to Prioritize and Communicate Risks | Matt Brown | 40m |
Customer Reliability Engineering adapts SLOs and error budgets into a quantifiable prioritization process by computing expected cost = likelihood × impact… | ||
| Automatic Metric Screening for Service Diagnosis | Yu Chen | 19m |
An automated pipeline detects sudden metric changes via kernel density estimation (including Gaussian, Poisson and Beta kernels), converts per-instance anomaly… | ||
| Whispers in Chaos: Searching for Weak Signals in Incidents | J. Paul Reed | 40m |
Research on an Etsy Cyber Monday outage shows slow personalized pages caused by CDN cache misses triggered by an HTTP 400 from a closed store referenced in a… | ||
| Architecting a Technical Post Mortem | Will Gallego | 42m |
Shows a concrete post-mortem workflow used at Etsy: interview actors within two days, run a one-hour meeting scheduled as 5 minutes warmup, 35–40 minutes… | ||
| Your System Has Recovered from an Incident, but Have Your Developers? | Jaime Woo | 17m |
A 40-person straw poll finds 43% stressed or very stressed after incidents, with >50% reporting worse sleep and concentration and 60% worse mood; 45% say no… | ||
| The History of Fire Escapes | Tanya Reilly | 42m |
Argues that New York’s long arc of disasters — from the 1835 and tenement blazes to the Triangle Shirtwaist fire (146 dead) — shows fire escapes are a hacky… | ||
| Leaping from Mainframes to AWS: Technology Time Travel in the Government | Andy Brody, James Punteney | 30m |
Helps migrate a legacy CBP Global Entry site onto AWS using infrastructure as code, switching from six-month VM waits to minute-scale provisioning; launches… | ||
| Operational Excellence in April Fools’ Pranks: Being Funny Is Serious Work! | Thomas A. Limoncelli | 19m |
An operations playbook for April Fools pranks: deploy behind feature flags so you can flip off a broken prank in minutes, use browser JS to run dark launches… | ||
| nothing matches | ||