Size a Persistent Disk and see the IOPS and throughput its capacity earns before the instance vCPU limit clips it.
On pd-standard, pd-balanced and pd-ssd you do not buy IOPS, you buy GB and the IOPS come with them at a fixed rate per GB. The instance then applies its own ceiling, which steps up with vCPU count, so a 4-vCPU VM cannot use anything like the rating of a multi-terabyte pd-ssd volume. Undersized disks throttle databases in a way that looks like an application problem, and oversized ones bill for IOPS the VM can never issue — the two limits have to be sized together. Verify the per-GB prices for your region in the Google Cloud pricing calculator, since disk rates differ noticeably between regions.
Persistent Disk
read IOPS = min(GB × IOPS-per-GB for the type, the per-instance ceiling for that vCPU count); cost = GB × type price × disks + snapshot GB × snapshot price.
read IOPS = min(GB × IOPS-per-GB for the type, the per-instance ceiling for that vCPU count); cost = GB × type price × disks + snapshot GB × snapshot price. On pd-standard, pd-balanced and pd-ssd you do not buy IOPS, you buy GB and the IOPS come with them at a fixed rate per GB. The instance then applies its own ceiling, which steps up with vCPU count, so a 4-vCPU VM cannot use anything like the rating of a multi-terabyte pd-ssd volume.
Undersized disks throttle databases in a way that looks like an application problem, and oversized ones bill for IOPS the VM can never issue — the two limits have to be sized together. Verify the per-GB prices for your region in the Google Cloud pricing calculator, since disk rates differ noticeably between regions.
This calculator takes 12 inputs: Provisioned size per disk, Number of disks, Disk type, vCPUs on the attached instance, pd-standard price per GB, pd-balanced price per GB, pd-ssd price per GB, pd-extreme price per GB, pd-extreme provisioned IOPS, pd-extreme price per provisioned IOPS, Snapshot data stored, Snapshot price per GB. The pre-filled defaults are a realistic starting point — replace them with figures from your own environment for a result you can act on.
Because Persistent Disk performance is capped per instance as well as per disk. The instance ceiling rises in steps with vCPU count, so a larger machine shape unlocks more of the IOPS your disk size already entitles you to.