Running Microsoft SQL Server in the cloud has never been straightforward. Most organizations want the elasticity and operational agility of the cloud, but they struggle with the reality that SQL Server is an I/O‑intensive, license‑heavy, and HA‑sensitive workload. Google Cloud NetApp Volumes (GCNV) changes this equation — especially for customers who can adopt SQL Server Failover Cluster Instances (FCI) instead of Always On AG (AOAG). With FCI + shared block storage, organizations keep full DBA control while unlocking dramatic TCO gains.
1) The Core Problem: Storage Performance Forces Oversized VMs
In many cloud designs, VM size is chosen not for CPU, but to hit storage throughput and IOPS targets. For example, Hyperdisk Balanced requires a 32‑vCPU VM to achieve 100K IOPS / 1,600 MiB/s for SQL, while GCNV Flex delivers equivalent or better performance with only 8 vCPU because performance is decoupled from compute.
Figure 1. Architecture comparison: GCNV + FCI (left) vs Hyperdisk Balanced + AG (right).
2) How FCI on GCNV Shrinks the SQL Cost Stack
- Smaller VM size: moving from 32 vCPU to 8 vCPU yields ~75% compute cost reduction.
- SQL Standard instead of Enterprise: FCI does not require AG features, enabling lower per‑core licensing.
- No 2–4× replica storage tax: shared storage removes duplicate database copies across replicas.
3) TCO Example (us‑central1) for mixed OLTP workload
Hyperdisk Balanced + AG + SQL Enterprise totals $25,458/month (compute $23,571, storage $1,887). GCNV Flex + FCI + SQL Standard totals $14,148/month (compute $5,893, storage $8,255). Net savings: $11,310/month — a 44% TCO reduction in this example.
Figure 2. Mixed OLTP workload (<1600MiB/s throughput) monthly cost comparison.
4) Why Performance Stays High on Smaller VMs
GCNV provides up to 160,000 IOPS and multi‑GiB/s throughput with <1 ms latency, independent of VM size, so DBAs do not need multiple disks, per‑disk tuning, or multi‑writer workarounds. Hyperdisk performance, by contrast, depends on available CPU on the VM to drive I/O.
Figure 3. Value chain: shared storage → smaller VMs → SQL Standard → lower TCO.
5) When to Choose FCI vs AOAG
Choose FCI on GCNV when you want lowest TCO, instance‑level protection (jobs, logins, agent), and do not need read‑scale.
Choose AOAG when you need readable secondaries or multi‑DB geo read‑scale and accept higher licensing/compute/storage costs.
6) Practical Migration Notes
- Lift‑and‑shift from on‑prem SAN layouts maps directly to GCNV LUNs; keep data/log/tempdb separation.
- Right‑size with the cohort table: for this Medium OLTP mixed workload example, 8 vCPU with regional GCNV volume meets the 100K IOPS / 1,600 MiB/s target.
- Use regional volumes for HA; for cross‑region DR, use async replication and automate failover for ~minutes‑level RTO.
Conclusion
GCNV turns SQL Server into a storage‑led, not compute‑led, deployment: shared block storage enables FCI, smaller VMs, SQL Standard licensing, and a single storage footprint. In the Medium OLTP mixed workload example, that translates to roughly $11K/month savings (~44%) per HA pair while maintaining sub‑millisecond performance and operational simplicity.
Learn more https://www.netapp.com/media/157610-sb-4482-cut-sql-server-costs-google-cloud-netapp-volumes.pdf