High Availability vs Fault Tolerance#

Every resilience question starts here: how much downtime is acceptable? The exam expects you to distinguish two very different promises - and to hunt for single points of failure (SPOF).

The definitions#

	High Availability (HA)	Fault Tolerance (FT)
Goal	Minimize downtime	Zero downtime
Behavior on failure	Recover quickly (seconds-minutes)	Continue without interruption
Cost	Moderate	Expensive (everything duplicated)
Example	Multi-AZ RDS (Relational Database Service) with failover	Aircraft systems, active-active multi-region

Reliability is the umbrella term: the system does what it should over time. Scalability (vertical vs horizontal) and recoverability feed into it.

Not: soruda “kabul edilebilir kesinti kisa olsun” varsa HA, “hic kesinti olmasin, hicbir kayip olmasin” varsa FT. “En ucuz olani” sorularinda ise genelde HA + iyi bir DR (disaster recovery) plani kazanir.

Where failures happen - the checklist#

Run every architecture through this list:

Instance dies → Auto Scaling Group replaces it (cattle, not pets - see auto-scaling-groups)
AZ (Availability Zone) dies → spread across 2+ AZs (subnets, ELB (Elastic Load Balancing), RDS Multi-AZ)
Region dies → multi-region: Aurora Global, S3 (Simple Storage Service) CRR, Route 53 failover
Dependency dies or slows down → decouple with SQS (see decoupling-with-sqs-sns)
Traffic spike → ASG (Auto Scaling Group) + ELB scaling
Deployment goes bad → multiple targets, rolling deployments, health checks

flowchart TB
    rt53[Route 53] --> alb[ALB across 2 AZs]
    alb --> ec2a[EC2 AZ-a]
    alb --> ec2b[EC2 AZ-b]
    ec2a & ec2b --> rds[RDS Multi-AZ]
    ec2a & ec2b --> sqs[SQS buffer]

Decoupling is resilience#

A synchronous chain (web → app → queue-less worker) fails as a whole. Putting SQS between tiers means the producer keeps working even when the consumer is down - the messages simply wait. Stateless compute + managed state (RDS/S3/SQS) is the exam’s golden architecture.

Exam check#

Q1: An architecture must tolerate the failure of an entire Availability Zone with no manual intervention and the least operational overhead. What is the minimum design?

A) Two EC2 (Elastic Compute Cloud) instances in the same AZ with an ELB
B) Instances across two AZs behind an ELB (or an ASG with multi-AZ subnets)
C) An EBS (Elastic Block Store) snapshot policy
D) A larger instance type in one AZ

Cevabı göster

Answer: B. Multi-AZ + ELB/ASG. A survives instance failure only, C is recovery not tolerance.

Q2: What is a “single point of failure” in a three-tier web app where all instances sit in one AZ?

A) The ELB
B) The AZ itself
C) The IAM (Identity and Access Management) role
D) The AMI (Amazon Machine Image)

Cevabı göster

Answer: B. Everything shares one failure domain - the AZ.

Ozet#

HA = kesintiyi kisalt (multi-AZ), FT = kesintiyi sifirla (aktif-aktif, pahali)
SPOF avla: instance, AZ, bolge, bagimlilik, trafik
Multi-AZ = dayaniklilik, multi-bolge = felaket kurtarma
Stateless + yonetilen servisler = sinavin altin mimarisi
SQS = tuketici olse bile ureticiyi yasatir (decoupling = resilience)