Multi-tenant Isolation
Robust Multi-Tenant Isolation for Secure AI Workloads
When you are running multiple AI workloads on shared infrastructure, you need them to have their own private compute and storage resources. You can do this by creating a Hedgehog VPC for each workload then attach resources to each VPC. A Hedgehog VPC is similar to a public cloud VPC. It provides an isolated private network with support for multiple subnets, each with user-defined VLANs and optional DHCP services.
Subnets can be isolated and restricted, with the ability to define permit lists to allow communication between specific isolated subnets. The permit list is applied on top of the isolated flag and doesn't affect VPC peering.
Isolated subnet means that the subnet has no connectivity with other subnets within the VPC, but it could still be allowed by permit lists.
Restricted subnet means that all hosts in the subnet are isolated from each other within the subnet.
A Permit list contains a list. Every element of the list is a set of subnets that can communicate with each other.
Cloud Native API
Modern Architecture for Scalable, Automated AI Networking
The Hedgehog AI Network is built on Kubernetes, which means our API is cloud native. We built Hedgehog on principles like scalability, automation, resilience, and microservices architecture.
Designed for Cloud Infrastructure
Operates seamlessly in containerized and orchestrated platforms like Kubernetes.
Stateless
All fabric configuration, topology, and operational state is declaratively managed through Kubernetes CRDs, making the system resilient to pod restarts and enabling GitOps workflows.
Microservices-Oriented
Built as modular, independent services that communicate via lightweight protocols (HTTP/REST, gRPC), enabling flexible deployment and scaling.
Automated and CI/CD-Friendly
Supports DevOps practices, including CI/CD pipelines, infrastructure as code (IaC), automated testing, and GitOps.
Resilient and Fault-Tolerant
Designed with failover, load balancing, and observability in mind.
Operates Like AWS VPC
Optimized for Private and Hybrid AI Clouds
Every AWS user on the planet uses VPC whenever they spin up their very first EC2 instance. VPCs exist in AWS regions and availability zones, and they provide subnets that control tenant access to EC2 instances.
Hedgehog VPCs exists in data centers for AI training, fine-tuning or inference at the data edge. They also provide subnets that control access to GPU, compute and storage resources.
We designed the Hedgehog VPC on the same principles as Amazon VPC, which is pretty much the same thing that Microsoft, Google and Oracle did for their cloud services. The difference is that Hedgehog VPCs are built for private and hybrid AI cloud use cases.
Hedgehog VPC Features
- Enables secure multi-tenant isolation for compute and storage resources
- Operates like AWS VPC
- Kubernetes-native API supports cloud-native toolchain
- Infrastructure as code
- Supports GitOps
- Zero touch provisioning
- Full life cycle management
- Includes Grafana, Loki, Prometheus observability
- Models network as Kubernetes cluster
- Edge fabric for AI inferencing
- GPU fabric for AI training
- Data Center fabric for core workloads
