Google Scale Abstractions
Build practical intuition for the canonical Google-era abstractions behind large-scale storage and data processing: GFS, MapReduce, Bigtable, and the design tradeoffs they made visible. Designed for engineers and data-platform builders who already know distributed-systems basics and want sharper abstraction judgment.
3 sections ยท 10 lessons
Course outline
Scale Forces
- Workloads Outgrow Machines
- Commodity Failure Budget
- Locality Beats Bandwidth
- Design a Failure Budget
Batch Stack
- GFS Shares Disks
- MapReduce Hides Recovery
- Movement Shapes Jobs
- Stragglers Set Tails
- Diagnose a Slow Job
Structured Storage
- Bigtable Sparse Map
- Tablets Move Load
- Beyond One Cluster
- Choose the Next Abstraction