Some very non-obvious lessons about how the cloud works:
Summary
Slava Akhmechet argues cloud providers function as collections of ~500 independent, hypergrowth startups governed by power laws and recursive dependencies.
Cloud services act as independent startups.
Approximately 500 services operate under one roof, with many following a power law where some are zombies and others reach billions in scale.
each more or less its own independent business
Recursive dependencies cause stack-wide failures.
Services like queues depend on compute, storage, and networking, meaning any single break propagates upward through the entire stack.
if anything breaks, everything up the stack breaks
Regional distribution ensures constant breakage.
Because services span dozens of regions, something is always broken somewhere, impacting everything positioned above the failure point.
something is always broken somewhere
Capacity constraints remain permanent.
Despite constant hardware buildouts, everyone faces capacity constraints, sometimes preventing important customers from acquiring resources at any price.
Everyone is always capacity constrained
Ecosystems provide tailwinds and risks.
Customers buy into the whole ecosystem rather than individual services, granting services free momentum but making the entire system vulnerable to a single faltering service.
each service gets free tailwind
Read the original ↗Save to your Sponge
Summarized with Sponge. Quotes are the author’s own words.