Mentora
Classytic

Course

Cloud Infrastructure: from one process to a reliable service

Build the portable mental model underneath AWS, Google Cloud and Hetzner: publish one controlled service endpoint, diagnose every boundary, distribute traffic only to healthy backends and scale before overload.

47 lessons4h 48mEnglish

What you'll learn

  • Identify the address, port, firewall and process involved at every service boundary
  • Explain a reverse proxy as a separate client-facing connection and upstream connection
  • Diagnose subnet, DNS, forwarding, firewall and listener failures by the last successful boundary
  • Compare round-robin and least-connections routing without treating either as new compute capacity
  • Remove an unhealthy target and predict how its lost capacity affects the remaining pool
  • Add healthy application instances until a finite request queue drains
  • Explain why application scaling can move the bottleneck to a shared dependency
  • Design a segmented VPC where only the service entry point is publicly reachable
  • Give a private application outbound NAT access without making it an inbound internet target
  • Explain why a matching route and an allowing firewall rule are both required
  • Choose ephemeral, block, object, file or database storage from workload semantics
  • Keep durable shared state outside replaceable application instances
  • Explain how edge cache hits reduce viewer latency and origin load
  • Choose cache lifetime while accounting for revalidation and stale content
  • Identify whether an instance, zone or region failure is inside the system’s designed resilience boundary
  • Keep service availability honest by placing required data across the same failure boundary as compute
  • Explain why multi-region resilience adds distributed-data and operational tradeoffs
  • Route a request near its customer while keeping market, currency, catalog and tax authority explicit
  • Use network location as a correctable guest hint rather than customer identity
  • Keep amount, currency, market price, tax and fulfilment inventory explicit through checkout
  • Select a recovery strategy from explicit RPO and RTO instead of the number of copies alone
  • Relate asynchronous replication lag to possible acknowledged-data loss
  • Explain why synchronous replication affects commit latency and why replicas do not replace backups
  • Describe fencing and promotion as separate requirements of safe database failover

Requirements

  • The Computer Networks course through “Running a network service,” or equivalent understanding of IPv4 subnets, DNS, TCP, TLS, NAT and ports

Course content

47 lessons · 4h 48m

About the creator

Classytic