Product Releases

VAST’s hyperscale vector index eliminates the traditional tradeoff between scale and performance by embedding a proprietary hierarchical clustering index directly into the VAST DataBase. Instead of relying on memory-heavy indexes or complex sharding that cause unpredictable latency, it organizes vectors into distance-based clusters and loads only the most relevant subsets during a search. This ensures predictable, memory-bounded query latency at a trillion-vector scale under a single governance model, making it ideal for large-scale RAG, fraud detection, and semantic search without the need for an external vector store.

VAST Native Compute brings managed Kubernetes directly into the VAST AI Operating System, allowing containerized applications, event-driven services, AI pipelines, and data processing workloads to run on dedicated VAST CNodes adjacent to the data they operate on.

Provisioning, scaling, upgrades, and lifecycle operations are handled entirely by the platform while compute gains direct access to VAST file, object, DataBase, and event services through a single operational model that reduces latency, simplifies operations, and eliminates unnecessary data movement.

Reversing the conventional approach of moving data to external execution engines, the VAST Native Query Engine brings analytical processing directly to where your data lives. With the 5.5 release, it natively supports over 50 new statistical and aggregation functions, such as variance, regression, and percentile estimation, alongside conditional filtering. This lets users to run complex statistical analysis, machine learning feature preparation, and hybrid vector-SQL analytics directly on the platform, entirely eliminating the latency, infrastructure sprawl, and governance gaps associated with data egress.

VAST 5.5 introduces a declarative, SQL-native partitioning model that accelerates large-scale analytics without adding operational complexity. Users simply define their partitioning intent (e.g., by time range or category), and the VAST engine automatically handles partition pruning, join optimizations, and efficient deletes at query time. This approach ensures high-performance scans and updates on massive datasets while completely eliminating the traditional administrative burdens of managing manual shards or rebalancing nodes.

VAST 5.5 extends workflow automation and reduces unnecessary data movement through two key capabilities. File-Triggered Pipelines bring real-time, event-driven automation to native NFSv4 workflows, allowing standard file creations or metadata changes to trigger pipelines instantly without requiring object storage re-architecture. Meanwhile, Instant Data Cloning for Block transforms heavy block-level copying (like VM clones) into zero-bandwidth metadata operations, allowing new targets to reference existing flash data instantly without draining system resources or storage bandwidth.