Storage-Optimized Z4D Machine Family Reaches GA

Z4D compute instances: what actually changed

Google Cloud has officially announced the general availability (GA) of its next-generation storage-optimized Z4D machine series in Google Compute Engine. Available as both virtual machines (VMs) and bare-metal instances, Z4D is specifically engineered for IO-intensive enterprise workloads, including large-scale relational databases, NoSQL engines like Cassandra and Aerospike, and real-time log analytics.

Powered by modern high-density NVMe SSD drives attached directly to the host hypervisor, Z4D instances deliver record-breaking random read and write IOPS while maintaining sub-millisecond latency under sustained saturation. The addition of bare-metal options allows enterprise clients to bypass hypervisor overhead completely for mission-critical core banking and telemetry systems.

Hardware architecture and IOPS benchmarks

The Z4D family features custom-tuned Intel Xeon processors paired with Google's custom Titanium infrastructure offload chips. By offloading networking, storage virtualization, and security policies to dedicated Titanium hardware, 100% of host CPU cores and memory channels remain available for guest application execution.

Benchmark tests shared by Google Cloud indicate up to 2.5x higher local storage throughput compared to previous-generation Z3 instances. The architecture supports non-disruptive host maintenance, ensuring high availability even during hypervisor firmware upgrades and hardware maintenance cycles.

Cost performance and enterprise deployment

For database administrators and cloud architects managing high-throughput clusters, Z4D offers a compelling price-to-performance ratio. By consolidating database nodes onto higher-density Z4D instances, organizations can reduce total VM count while satisfying strict IOPS and memory requirements.

Z4D instances are available today across major Google Cloud regions in North America, Europe, and Asia-Pacific, with full support for terraform provisioning and Google Kubernetes Engine (GKE) local-SSD node pools.