Spice.ai Enterprise
Spice.ai Enterprise documentation — self-hosted, production-grade deployment of the Spice runtime.
Spice.ai Enterprise is a self-hosted, production-grade distribution of the Spice.ai runtime for organizations that need enterprise security, high availability, and dedicated support.
Enterprise extends the open-source Spice runtime with:
- Multi-node clustering and HA — Multi-active distributed query with automatic failover, built on Apache Ballista.
- Distributed accelerations — Partition-sharded accelerated datasets across executors with write-through semantics. See Distributed Accelerations.
- Kubernetes Operator — Automated deployment, scaling, and lifecycle management via
SpicepodSetandSpicepodClusterCRDs. - Acceleration snapshots — Object-store-backed snapshot, bootstrap, and recovery for accelerated datasets. See Acceleration Snapshots.
- Authorization policy — Cedar-based fine-grained access control over datasets, models, tools, and endpoints.
- Enterprise distributions — NAS (SMB/NFS), CUDA GPU-accelerated, data-only, and ODBC connector builds.
- Authentication — OIDC bearer tokens, API keys, and combined auth with identity SQL functions.
- mTLS cluster security — Auto-provisioned certificates for secure inter-node communication.
- SOC 2 Type II compliance — Code and security audited.
- Tiered security updates — Up to 3 years of guaranteed patches.
- 24/7 premium support — Dedicated account team via Slack Connect, email, and on-call pager.
- 99.9%+ SLA — Enterprise uptime guarantee with on-call support.
Deployment Options
| Option | Description |
|---|---|
| Kubernetes (Helm) | Deploy using the Spice.ai Kubernetes Operator with Helm charts |
| Docker | Enterprise container images from the AWS Marketplace ECR |
| AWS Marketplace | BYOL and Plan-based licensing via AWS Marketplace |
| AWS AMI | Pre-built Amazon Machine Images for EC2 deployments |
Enterprise vs Open Source vs Cloud
| Open Source | Cloud | Enterprise | |
|---|---|---|---|
| Pricing | Free (Apache 2.0) | Subscription + usage-based | Enterprise licensing |
| Hosting | Self-hosted | Fully managed | Self-hosted BYOL |
| Scale | Single-node | Multi-node, clustering, HA | Multi-node, clustering, HA |
| Control Plane | Static, YAML-driven | Dynamic, remote (Portal/API) | Dynamic, remote/local (API/YAML) |
| Distributions | Default + Metal | All | All (incl. NAS, ODBC) |
| Security Updates | Latest minus-one (6-8 weeks) | Rolling, automated | Tiered, up to 3 years |
| SLA | N/A | 99.9% | 99.9%+ |
| Support | Community | Standard (email, Slack) | Premium 24/7 on-call |
| Compliance | N/A | SOC 2 Type II | SOC 2 Type II |
Production-Ready by Design
Spice.ai Enterprise is the supported path for running Spice in production. The Production Readiness section is the canonical reference for promoting a deployment to production:
- Production Readiness Checklist \u2014 sign-off list covering topology, storage, observability, security, and upgrades.
- High Availability \u2014 multi-replica, multi-AZ, and
SpicepodClustertopologies. - Storage \u2014 NVMe,
io2Block Express, Premium SSD v2, and acceleration sizing. - Observability \u2014 Prometheus metrics, Grafana dashboard, log routing, and the recommended alert set.
- Security \u2014 image pinning, Pod Security Standards, NetworkPolicy, mTLS, and secrets management.
- Upgrades \u2014 versioning, rolling upgrades, and rollback procedures.
Get Started
- QuickstartDeploy Spice.ai Enterprise on Kubernetes in minutes.
- DistributionsChoose the right runtime distribution for your workload.
- Kubernetes OperatorManage Spicepods with the Spice K8s Operator.
- AuthenticationConfigure OIDC, API keys, and identity functions.