Filestore Service Tiers for the PCA Exam

GCP Study Hub
Ben Makansi
December 8, 2025

Filestore is one of those services where Google's documentation will hand you a wall of numbers. IOPS at every capacity step, throughput per terabyte, minimum and maximum sizes for each tier, regional availability quirks. Memorizing all of that is a waste of time for the Professional Cloud Architect exam. The exam does not test you on whether you know that the Zonal tier starts at a particular IOPS number. It tests whether you can pick the right tier given a workload description.

Two attributes do almost all of the work: whether performance scales automatically with capacity, and whether the file share survives a zone failure. Map a workload onto those two attributes and the right tier falls out.

The Two Attributes That Matter

Filestore has four service tiers worth knowing for the exam: Basic, Zonal, Regional, and Enterprise. Rather than remembering each tier's full spec sheet, anchor on what makes them different from each other.

  • Automatic performance scaling means IOPS and throughput grow as you add capacity, without you having to provision or resize anything. Only the Enterprise tier offers this.
  • High availability across zones means the file share keeps serving reads and writes if one zone goes down. Regional and Enterprise both offer this. Basic and Zonal do not, because they live in a single zone.

Those two attributes are the lens. Every Filestore question on the Professional Cloud Architect exam can be answered by checking whether the workload needs one, both, or neither.

Rule One: Performance Scales With Storage Means Enterprise

If a question describes a workload where performance has to grow automatically as the dataset grows, the answer is Filestore Enterprise. Enterprise is the only tier where you do not have to predict or manually scale your performance ceiling. As you add capacity, IOPS and throughput scale with it.

The language to watch for in exam questions is anything about performance keeping up with data growth without operator intervention. Phrases like "performance must scale automatically with capacity," "IOPS need to grow as the dataset grows," or "the team does not want to manually resize for performance" all point at Enterprise. None of the other tiers handle this.

Rule Two: High Availability Across Zones Means Regional or Enterprise

If a question emphasizes uptime, cross-zone redundancy, or mission-critical availability, you need either Regional or Enterprise. Both replicate data across zones inside a region, so a zone outage does not take the file share offline. Basic and Zonal both live in a single zone and go down with that zone, which disqualifies them for any workload where zone-level high availability is a requirement.

The choice between Regional and Enterprise comes down to whether the workload also needs automatic performance scaling. If it does, Enterprise. If it does not, Regional gives you the high availability without the cost of the scaling feature on top.

Rule Three: No HA and No Auto-Scaling Means Zonal or Basic

If a workload tolerates some downtime and has predictable performance needs, you do not need to pay for high availability or automatic scaling. Zonal works for standard file sharing and dev workloads where a zone outage is acceptable. Basic is the cheapest tier and is appropriate for pure dev and test environments where cost is the dominant factor.

The exam does not usually force a choice between Zonal and Basic at a fine grain. If a question is pushing you toward a single-zone tier, the distractors will typically be Regional or Enterprise rather than splitting hairs between Zonal and Basic. Pick the single-zone option and move on.

Putting the Rules Together

The decision tree collapses to three checks. Does the workload need performance to scale automatically with capacity? If yes, Enterprise. Does the workload need to survive a zone failure? If yes and you also need automatic scaling, Enterprise. If yes and you do not need automatic scaling, Regional. If neither is required, Zonal for general use or Basic for the cheapest dev and test workloads.

That is the entire framework I use for Filestore questions on the Professional Cloud Architect exam. The numeric specs in the documentation are interesting if you are sizing a real production deployment, but they are not what the exam is checking. The exam is checking whether you can match a workload's requirements to a tier, and those two attributes are the only ones that matter for that match.

One distractor pattern worth flagging: questions that mention performance requirements without saying anything about automatic scaling. If a workload just needs a certain IOPS or throughput level and that level is fixed, you do not automatically need Enterprise. Regional or Zonal at an appropriate capacity may meet the spec. Enterprise is specifically for the case where performance has to grow with data without you intervening.

My Professional Cloud Architect course covers Filestore service tiers alongside the rest of the advanced architecture material.

arrow