Availability is the ability of a system to remain operational and accessible during failures or disruptions, whether that's a zonal outage, a regional fiber cut, or a hardware failure inside a single rack. On the Professional Cloud Architect exam, when a question states that the business requires "high availability," you are being asked to pick the service or configuration that minimizes downtime under the failure mode the question describes.
That usually means choosing between options like a single zonal Compute Engine VM, a regional managed instance group, a multi-region Cloud Spanner instance, or a Cloud SQL instance with a failover replica. The right answer is the one whose redundancy model matches the failure the question is testing.
Availability is expressed as an uptime percentage over a year. The higher the percentage, the less downtime the system is permitted. The numbers you should know cold for the exam:
Each additional nine cuts the allowed downtime by an order of magnitude, which is why moving from four nines to five nines is a much bigger architectural commitment than the percentages suggest.
A Service Level Agreement is the formal guarantee Google Cloud publishes for a given service at a given configuration. It defines the maximum allowable downtime and the financial credits Google owes you if the service falls below that target. SLAs are configuration-specific. A zonal Compute Engine VM has a different SLA than a regional managed instance group, and a regional Cloud SQL instance has a different SLA than a Cloud SQL instance with cross-region replicas.
For the Professional Cloud Architect exam, the SLA tier is often what separates two otherwise plausible answers. If the scenario calls for five nines and one option is a zonal deployment, that option is wrong regardless of how clean the rest of the architecture looks.
Cloud Spanner in its multi-region configuration carries a 99.999% availability SLA globally. That is why Spanner shows up in exam scenarios involving global financial systems, payment processing, or any workload where five minutes of yearly downtime is the ceiling. If the question describes a relational workload that needs strong consistency and five nines across regions, Spanner is the answer the exam is looking for.
Compare that to a regional Cloud SQL instance with a high-availability configuration, which lands at 99.95%. Both are managed relational databases, but their availability tiers are an order of magnitude apart, and the exam expects you to know which one fits which requirement.
Downtime affects business continuity and user satisfaction directly. Mission-critical systems like healthcare records, trading platforms, and core banking ledgers cannot tolerate the 1.83 days a year that 99.5% allows. When a Professional Cloud Architect question frames a system as mission-critical, the implicit floor is usually four nines, and sometimes five. Reading the scenario for those cues is half the work of answering availability questions correctly.
My Professional Cloud Architect course covers availability and SLA tiers alongside the rest of the foundational architecture material.
