Solutions

Applies to VAST Cluster 5.5.

Solutions are VAST's plays for a business goal, such as an AI factory, a cyber lakehouse or a neocloud. Each one below links to VAST's own page for it, then lists what to read in these docs to build it, in order. Most steps are end-to-end patterns: step-by-step recipes for combining two or three services.

Enterprise AI Factory

Run AI pipelines next to your data: ingest documents and media, embed them, retrieve context and serve it to models, on one platform. VAST: Build an Enterprise AI Factory at Scale.

In these docs:

1. Retrieval for RAG, with permissions: chunks, embeddings and permission-aware retrieval.
2. Process new objects into a table: run a function on every new object.
3. VAST Vector Search and VAST DataEngine: the two services underneath.
4. Reference applications: a RAG research assistant and a video search application built this way.

Cyber lakehouse

Keep security logs and their long-term history on VAST, and query them with SQL, alongside the tools you use for real-time alerting. VAST: VAST for Cybersecurity (solution brief), the next-generation cyber lakehouse (white paper), Splunk.

In these docs:
Logs can come in two ways, and you can use either or both:

1. Stream JSON events into a table: log events through a Kafka topic, with blob expansion turning each JSON event into a typed row as it arrives.
2. Land files and query them with SQL: log files, batch or historical, landed as Parquet and imported into tables.
3. Trino and Spark: ad hoc and large-scale queries over the tables.
4. Find and audit data: who accessed what on the cluster itself.

Not covered in these docs: shipping logs into VAST from a log pipeline or a SIEM.

Data analytics and lakehouse

Run SQL and Spark on tables and files that live on one platform, with the query engine of your choice. VAST: Data Analytics, with Spark, Starburst, Dremio and Vertica.

In these docs:

1. Land files and query them with SQL: import Parquet into tables.
2. Data modeling and querying: table design and the query engines.
3. Trino and Spark: connect the engines.
4. Snapshots and replication: point-in-time copies and disaster recovery for your tables.

Real-time event analytics

Query streaming events with SQL as they arrive, without a job that copies them into tables. VAST: Real-Time Event Analytics.

In these docs:

1. Stream JSON events into a table: events as typed rows, in the same transaction as the produce.
2. Query topics as tables: read a topic's raw events with SQL.
3. Event sourcing and CQRS: an event log and a current-state table built from it.
4. VAST Event Broker: the service underneath.

Neoclouds

Run many customers' AI workloads on one VAST cluster, each in its own tenant with its own network, users, keys, capacity and performance limits. VAST: Building AI Factories: a blueprint for next-gen GPU clouds (blog).

This path is for developers who build on a neocloud tenant, or who build the provider's own tenant-facing services. Creating and administering tenants is a cluster admin's job, in the KB: creating tenants and managing tenant administrators.

In these docs:

1. Tenants and one Element Store: what a tenant owns and what it shares with the rest of the cluster.
2. Developer tenant: a tenant admin sets up users, keys, databases, buckets and Kafka views inside one tenant. The same steps work for any tenant.
3. Quotas and quality of service: the capacity and performance limits a tenant runs under.
4. Provision and deploy from CI and automation tools: create each tenant's objects from code.
5. Enterprise AI Factory: the workloads tenants run on top, and the VAST Catalog access page for what each tenant can see in the Catalog.

See also

• End-to-end patterns: the recipes these solutions are built from.
• Choose a service: start from what you want to do.