Part 0 · 1 chapters · ~10 min

IaaS, PaaS, SaaS, FaaS

The four ways to buy computing, what each hands to the provider and what it leaves you, and how to decide what to build and what to buy, with a fintech worked through capability by capability.

1

The layers, and where responsibility shifts

four ways to buy computing
  1. IaaS: virtual hardware. You run everything above it.
  2. CaaS and managed Kubernetes: a standard way to deploy every service.
  3. PaaS: push code or an image, and the platform does the rest within its opinions.
  4. FaaS: a handler per event, for glue and spiky workloads.
  5. SaaS: a finished capability whose outages and pricing become yours.
  6. The decision: build what differentiates you and buy the rest at the highest layer your constraints allow.
capability (a fintech)build or buylayerwhy
ledger and balancesbuildcontainers on k8s + managed Postgresthe core of the product; correctness and control matter most
risk and fraud rulesbuildcontainers + a streamdifferentiating, data-specific
BVN / NIN verificationbuySaaS (a KYC provider)regulated integrations someone else maintains
card issuingbuySaaS (an issuing processor)scheme certification is years of work
SMS, email, pushbuySaaSdeliverability is a specialism
marketing sitebuyPaaS (a frontend platform) + headless CMSno reason to run servers for it
statement PDF generationbuild, smallFaaS on a queuespiky, monthly, stateless
observabilitybuySaaS or the cloud's native stackrunning Prometheus at scale is its own team
the lock-in question
Every layer you buy is a dependency you cannot fix yourself. Keep the boundary narrow, behind an adapter you own (the Platformization course part 5), keep your data exportable, and know your exit cost before you need it. Lock-in is a price, and paying it is often worthwhile.
IAAS, PAAS, SAAS, FAAS
four ways to buy computing, each handing a different slice of the stack and of the pager to the provider
swipe the figure sideways, or tap expand for full screen
1/6
IaaS
IaaS (EC2, Compute Engine, block storage, VPCs): you rent VMs, disks and networks, and build everything above. Maximum control: any OS, any runtime, any network design. Maximum work: patching, scaling, backups, monitoring of everything you install. Teams that choose IaaS are choosing to run an operations function.