Vast Data expands Google Cloud partnership for AI platform

Vast Data has expanded its partnership with Google Cloud, allowing customers to use its AI operating system as a managed service across hybrid cloud and on-premises environments.

The platform now enables enterprises to connect clusters in different locations through a unified global namespace, demonstrated by linking clusters between the US and Japan for near real-time data access.

The system can also connect directly to Google Cloud’s Tensor Processing Unit virtual machines, with model load speeds similar to local NVMe disks in tests using Meta’s Llama-3.1-8B-Instruct model.

The platform is available through Google Cloud Marketplace, with reference guidance provided for eligible customers and partners.

🔗 Source: Vast Data

🧠 Food for thought

Implications, context, and why it matters.

Speed gains meet real-world infrastructure challenges

Near-real-time data access between US–Japan clusters sounds strong, yet without published Service Level Agreements or latency benchmarks from Vast Data (VAST) in 1 teams cannot judge reliability for production AI workloads.
Marketing touts Non-Volatile Memory express (NVMe)-class performance when linking data to Google Cloud Tensor Processing Units (TPUs) and GPUs 1, but proof at 70B+ parameter scale needs third-party benchmarks.
Unified global namespace gives one logical view and decouples compute from storage for independent scaling 2. Cross-region transfers between the US and Japan can trigger Google Cloud egress fees, with inter-region pricing that varies by route and destination 3.

FinOps teams can use hybrid cloud cost transparency gaps

Cross-border AI transfers like US–Japan face Google Cloud egress fees that vary by destination and network path 3. FinOps firms can model these costs and plan Cloud Interconnect to cut spend for enterprises using VAST’s hybrid setup.
Google Cloud egress pricing changed in February 2024 for Premium Tier internet egress and inter-region transfers 4. Providers who understand the updated rates and interconnect discount models 3 can tailor cost reviews for VAST’s multi-region AI rollouts.
Network optimization vendors can target enterprises on VAST’s federated AI stack under one namespace. Pair Console Connect’s Layer 3 CloudRouter, a managed Layer 3 IP routing service, for dynamic routing 5 with smart caching to limit data moved to Google Cloud TPUs from on-premises clusters.