Collision 2015 - Urs Holzle of Google with Bruce Upbin of Forbes
summary
Google runs its own global infrastructure with billions of dollars a year in capex and opex and chooses custom hardware, containers and cluster OSes over plain VMs. App Engine powers apps like Snapchat; YouTube backend gets an asserted 80% cost cut early on. Data processing moves from MapReduce to Cloud Dataflow / MillWheel, letting 100-line jobs replace much larger MapReduce code. Public clouds host roughly 1% of server cycles today, so most enterprise workloads remain off-cloud.
AI-written summary · about
chapters (10)
- 0:00The scale of Google infrastructure
- 1:34Early days and decision-making
- 2:48The evolution of App Engine
- 4:00Overcoming survival risks
- 5:16The strategic acquisition of YouTube
- 7:02Culture and long-term innovation
- 9:09Data center efficiency and cloud adoption
- 12:20Technological shifts in the cloud
- 16:02The changing IT landscape
- 18:24Startups and infrastructure growth
related talks
13mGoogle's Cloud Computing Vision
11mAbstracting infrastructure away from developers
25mAn Introduction To Google Cloud Platform
02h 15mGoogle Cloud Next '17 - Day 1 Keynote
17mScott Raney | Google Next 2017
36mIntroduction to High Performance Computing on Google Cloud Platform
04mLightning Talk: Google Cloud Data Services and Cloud Foundry
57mDebugging Java in Production on Google Cloud Platform