Instiq
Chapter 7 · System architecture·v1.0.0·Updated 7/7/2026·~11 min

What's changed: Initial version (topic 2.13, subtopics 2.13.1–2.13.4)

7.3System Architecture on Cloud Services

Key points

Learn how to architect systems on IaaS (Infrastructure as a Service): the difference between ephemeral storage and persistent storage, fixed IP vs floating IP, isolation via tenant networks, traffic control with firewalls and security groups, object storage, messaging systems/queues, and automatic node adjustment via an autoscaler.

Unlike an in-house data center, IaaS offers servers, storage, and networking as resources provisioned through an API. The HA and scalability ideas from earlier sections still apply, but understanding cloud-specific building blocks—ephemeral storage, tenant isolation, autoscalers—matters both in practice and on the exam.

7.3.1Storage and IP address types

  • Ephemeral storage is tied to an instance's lifecycle: terminate the instance and the data disappears. Use it for things you can rebuild (the OS, caches). Persistent storage has a lifecycle independent of any instance (e.g., a block-storage volume) and survives instance termination—the right place for data you cannot afford to lose, like database files.
  • Fixed IP is an IP address permanently assigned to an instance. Floating IP is an address reserved independently of any instance and reattached to whichever node needs it—reattaching it to a healthy node on failure achieves failover without changing the IP address clients use.

7.3.2Network isolation and supporting services

  • A tenant network gives each user (tenant) a logically isolated network space while sharing the cloud provider's physical infrastructure—traffic from other tenants is invisible and never mixes with yours. On top of that, firewalls and security groups (permit rules applied per instance, typically stateful so return traffic for an allowed connection passes automatically) restrict traffic to only what is needed.
  • Object storage handles files as key-value objects accessed over an HTTP API (suited to large volumes of images, backups, logs; unlike block storage, it is not mounted as a filesystem).
  • Messaging systems/queues mediate service-to-service communication asynchronously: the sender just places a message on the queue, decoupling it from the receiver's processing speed or temporary outages. An autoscaler automatically adds or removes instances based on metrics like CPU utilization—cloud automation of the scale-out/in idea from the previous section.
Exam point

The most common contrast: ephemeral storage disappears when the instance is terminated; persistent storage survives independently of the instance. Also common: reattaching a floating IP achieves failover without changing the client-facing address. Expect questions on security groups being per-instance and typically stateful (return traffic auto-permitted), and on object storage being key-value over HTTP, not something you mount as a filesystem.

The key discipline in IaaS design is always asking "what disappears, and what survives?" An autoscaler adds and removes web-server instances based on load, but if DB data or user-uploaded images are stored directly on those instances' ephemeral storage, the data vanishes the instant a scale-in event terminates one. The standard division of labor is therefore: keep the database on persistent storage (a dedicated instance or managed service), and keep user uploads in object storage. For failover, designing around a floating IP lets you reattach just the address from a failed node to a healthy one—traffic switches instantly with no DNS-propagation wait (with only fixed IPs, swapping out a whole node means reconfiguring the address itself). On the networking side, because a cloud shares physical infrastructure across many customers, tenant network isolation is the baseline, and inside that you layer security groups for least-privilege rules—"web servers: only 80/443," "DB server: only port 3306 from web servers." If services communicate only via synchronous HTTP calls, one service stalling or going down drags the whole chain to a halt; inserting a messaging system/queue to go asynchronous means messages simply pile up in the queue while the receiver is briefly unavailable, and the sender keeps working regardless—a real gain in resiliency.

ConceptCharacteristicTypical use
Ephemeral storageDisappears when instance is terminatedOS, temporary cache
Persistent storageSurvives independently of the instanceDatabase data
Fixed IPPermanently bound to an instanceOrdinary server communication
Floating IPReassignable to another nodeFailover without changing the IP
Warning

Trap: "ephemeral storage is safe for DB data as long as you never manually delete the instance" ignores that scale-in or unintended re-creation can terminate an instance anyway—data you cannot afford to lose belongs on persistent storage. Also wrong: "a fixed IP can be instantly reattached from a failed node to a healthy one"—reattachment is exactly what floating IP is for; a fixed IP is, by definition, bound to one instance and was never designed to move.

Ephemeral vs. persistent storage, fixed vs. floating IP, and security groups within a tenant network.
Always ask: does it disappear or survive?

7.3.3Section summary

  • Ephemeral = tied to the instance's fate / persistent = survives independently. Fixed IP = permanent / floating IP = reattachable for IP-stable failover
  • Tenant network + security groups give logical isolation and least-privilege traffic. Object storage, queues (async decoupling), and autoscalers are the standard cloud building blocks

Sign in to track progress — Log in.

Quick check

(just a quick review)

Q1. Which storage type loses its data as soon as the instance is terminated?

Q2. You want to fail over from a failed node to a healthy one without changing the client-facing IP address. Which mechanism should you use?

Q3. Service A relies on synchronous HTTP calls to service B; when B becomes temporarily unresponsive, A also stalls. Which mechanism helps loosen this coupling?

Check your understandingPractice questions for Chapter 7: System architecture

Keep track of your progress

The full study guide is free to read. Sign up free to practice with the question bank, track what you have read, review your mistakes, and highlight passages.