The GCP architecture choices that cut cost

Network tier: premium versus standard

GCP offers two network service tiers. Premium routes egress over Google's backbone for lower latency and global reach; standard hands traffic to the public internet sooner at a lower per GB rate. For workloads where a few milliseconds do not change the user experience, batch egress, internal tooling, or regional traffic, standard tier can cut egress materially.

The decision is per workload, not global. Keep premium where latency is part of the product and move the rest to standard.

Storage location and class

Cloud Storage prices regional, dual region, and multi region buckets differently, and the storage classes standard, nearline, coldline, and archive trade cheaper storage for higher access cost and minimum durations. Data placed in multi region for resilience it does not need, or kept in standard when it is read once a quarter, overpays on both axes.

Match location to the resilience the data actually requires and class to how often it is read. Sustained use discounts, which apply automatically to eligible Compute Engine usage, then reward steady workloads without any commitment.

BigQuery and the compute model

BigQuery bills two ways: on demand by bytes scanned, or capacity with reserved or autoscaling slots. Spiky exploratory query patterns often suit on demand, while steady high volume workloads usually cost less on capacity pricing. Choosing the wrong mode can double the bill for the same queries.

For services, Cloud Run scales to zero and bills per request and resource, which suits bursty or low duty cycle workloads, while GKE earns its keep on steady, dense, always on services. Running a low traffic service on an always on cluster pays for idle nodes; running a heavy steady service serverless pays a premium per request.

A worked comparison

Indicative figures, verified against the client's billing data, anonymized. A European SaaS company realigned three architecture choices.

Your next step

Review each workload against the pricing model it is using and move the mismatches. For the full method read the GCP cost optimization guide, and for neighbouring detail see network tiers premium versus standard and your first GCP cost optimization sprint. To apply it, our GCP cost optimization service turns the review into verified savings, and you can request a free trial.