Embracing Serverless Ops
summary
Argues serverless shifts ops into developer hands while still needing ops tooling: IOpipe + Amazon Lambda + Kinesis replace VMs and Elastic Beanstalk for ingest, with API Gateway as a cheaper front end; notes concrete limits—five minute max duration, memory tiers from 128MB up to 1.5 gigabytes, default 1,000 concurrent containers—and recommends monitoring for cold starts and memory leaks. Describes building a Slack image bot that misses Slack’s ~2s response window until profiling with a V8 profiler exposes slow disk writes; webpack cuts lazy-loads by hundreds of milliseconds.
AI-written summary · about
chapters (26)
- 0:00Introduction
- 0:57What is Serverless
- 2:02Serverless and DevOps
- 3:18Service
- 3:55Continuous Deployment
- 7:42The 12 Factor App
- 8:49Soleus Manifesto
- 11:15Converged Tools
- 12:04Developer Tools
- 12:36Build slowly
- 13:37Amazon Kinesis
- 15:36API Gateway
- 17:51Stealth Force Art
- 18:15Lambda
- 18:36Hello World
- 19:10Memory
- 20:15Containers
- 22:39Memory leaks
- 24:13Ops still exist
- 24:26Profiler
- 25:18Profiling
- 25:48Slack Bot
- 26:56The Problem
- 27:20The Profiler
- 27:45Testing the Profiler
- 29:53VA Profiler
related talks
23mServerless Ops: What to Do / This is What We Do, When the Server Goes Away
28mThe Brave New World of serverless computing
58mServerless - the brief past, the bewildering present, and the beautiful (?) future
42mNo Server Is Better Than No Server: Building Blocks for Serverless Architectures
28mServerless - the way microservices should be
05m#Serverless + #Devops = BFF
28mFun With Serverless Javascript
37mTHE DEVOPS EVOLUTION