Cargo-cult hyperscale

Copying Netflix, Google, or Amazon patterns into a org whose scale problem is actually a spreadsheet and a single Postgres instance.

Cargo-cult hyperscale is what happens when conference talks become procurement requirements. An eighty-person B2B shop adopts Kubernetes, service mesh, and eventual consistency because a FAANG engineer said it on stage — while their real outage cause is an unindexed query and a manual deploy on Fridays.

Hyperscale patterns are real and load-bearing at hyperscale. They are also expensive: headcount, observability, coordination tax, and operational culture that mid-market companies do not have and cannot hire fast enough to fake. Copying the topology without copying the constraints is cosplay with an on-call rotation.

The historical arc runs from "we need Hadoop" for gigabytes to "we need microservices" for four teams to "we need Kafka" for twelve events a minute. Each time the infra bill and incident surface area grow faster than the product. Consolidation follows years later, dressed as "platform engineering" or "rationalization."

The antidote is insultingly simple: measure your bottleneck. If it is not network fan-out or multi-region write throughput, you probably need better tests and a modular monolith, not a mesh. Fashion hates that answer because it does not sell certificates.

Exhibits

Architecture · 2012–now

Microservices Everywhere

Costly fad

Independent deployability as religion. Correct for some; ruinous cargo-cult for most.

$ Teams split 4-person products into 40 services, then hired SREs to babysit the mesh. Many later consolidated — after years of distributed tracing invoices and on-call trauma. See case studies on premature microservices and big-bang cutovers for named failure modes.

Platform · 2014–now

Kubernetes

Stuck

Container orchestration that became the cloud's OS. Overkill for many; unavoidable for many others.

Platform · 2011–now

Apache Kafka

Stuck

Durable log as nervous system. Event-driven architecture with real operational teeth.

Platform · 2006–2018

Hadoop Everywhere

Costly fad

MapReduce as lifestyle. The elephant in rooms that needed a spreadsheet.

$ Enterprises stood up Hadoop clusters for gigabytes of data that fit on one Postgres instance. Hadoop admins, ZooKeeper nightmares, and ETL rewrites burned millions before Spark and cloud warehouses retired the elephant.

Platform · 2009–2020

Apache Mesos

Faded

Datacenter kernel before Kubernetes won the orchestration war. Twitter survived on it; most did not need it.

Platform · 2013–now

Docker / Containers

Stuck

Ship the environment with the code. The packaging unit of modern compute.

Practice · 2009–2019

NoSQL for Everything

Costly fad

Schema-optional databases sold as a lifestyle. Right tool for some jobs; default for none.

$ Startups ditched Postgres for document stores "because scale," then rebuilt relational integrity in application code — and paid twice when joins came back as a product requirement.

Architecture · 1970–now

Modular monolith

Stuck

One deployable with clear boundaries. The unfashionable default that keeps shipping while microservices argue about ownership.

← All patterns

© 2026 Fadstack · Shane Code

Opinionated history · not a ranking