Skip to content

More stories

Tech & AI2 min read

Google Cloud Makes Storage-Optimized Z4M Machine Series Generally Available

Google Cloud launched its Z4M storage-optimized VM and bare-metal instances, offering up to 168,000 GiB of local SSD and 400 Gbps RDMA for AI/ML workloads.

Loading...

Google Cloud (GOOGL) announced the general availability of its next-generation Storage-optimized Z4M machine series in Compute Engine, offered as both virtual machines and bare-metal instances.

The company built Z4M for AI/ML training and inference workloads that handle huge data volumes, such as checkpointing, distributed key-value cache and parallel file systems. The series provides up to 168,000 GiB of local SSD storage paired with up to 400 Gbps of Cloud RDMA networking. RDMA reduces data-access latency by up to 60% compared with TCP, which helps keep GPUs and TPUs fully utilized while right-sizing cloud investment.

Z4M VMs and bare-metal instances offer up to 192 vCPUs and up to 1,536 GiB of memory, with 875 GiB of local SSD per vCPU. Storage runs on Titanium SSDs, which offload local storage processing from CPU resources. Compared with the previous Z3 family, Z4M delivers up to 15.6 million random-read IOPS and up to 75,600 MiB/s sequential-read throughput, improving LSSD storage performance by up to 70% and throughput by up to 110%.

The 400 Gbps RDMA network supports NVIDIA GPUDirect Storage and TPUDirect Storage, transferring data at line rate with a balanced LSSD-to-networking throughput ratio of 1.4:1. Bare-metal instances are positioned for agentic AI workloads, including isolated sandboxes at scale with Agent Substrate.

Partner testing reported in the announcement includes Silk measuring up to 3.7x lower latency and up to 6x lower CPU usage with Cloud RDMA versus TCP, and WEKA measuring roughly twice the read throughput and up to 79% more IOPS against the Z3 family. Nutanix, Sycomp and others also provided statements.

For fleet management, Cluster Director can configure and deploy large Z4M clusters, with compact or spread topologies for availability domains. Maintenance gives several days of notice; VMs with up to 42,000 GiB of LSSD can live-migrate, while larger VMs and bare-metal instances must be terminated and restarted while preserving data. Z4M also supports Hyperdisk, with up to 512 TiB of capacity per instance.

Z4M VMs and bare-metal instances are available today in select regions worldwide, selectable under the Storage-Optimized machine family in the Google Cloud console.

Alphabet shares were up 0.78% in delayed trading.