Just enough DevOps for data scientists.
summary
Practical checklist for running Salesforce Einstein at scale: monitor p99 latency and request/error KPIs on a simple dashboard, collect host metrics with StatsD, and plan for failure and blue‑green deploys. Enforce auditable changes with tools like Shaper schema auditor, use a config manager (Consul/Puppet), apply IAM roles at user/service/job levels, track per‑host resource allocation so Spark/EMR containers actually launch, and auto‑shutdown idle instances with Janitor Monkey to control costs.
AI-written summary · about
chapters (13)
- 0:00Intro
- 0:39About Anya
- 1:50What is DevOps?
- 2:45Fastest Growing Top 5 Enterprise Software Company
- 6:44Blue Green Deployments
- 7:24Assume people make mistakes
- 10:25Tip 4: Changes should be auditable Schaper the tool to compare schemas
- 11:41Configuration management
- 12:42Pick a naming convention
- 13:21Permissions
- 14:37Understand resource allocation
- 16:47Tip 9: Monitor multiple viewpoints Connectivity Viewer
- 18:44Getting started tips
related talks
20mHow AWS infrastructure addresses Salesforce’s resiliency goals
32mA Peek Inside the Enterprise Cloud at Salesforce
30mDevOps For Data Teams
15mDevOps at scale
35mScaling up to Big Data Devops for Data Science
47mMetrics Driven Machine Learning Development at Salesforce Einstein
15mOptimizing Salesforce for Scalable, Real-Time Performance
30mA Devops Approach to Data: A Tale Through the Trenches in Search of a Cloud Data Revolution