I’m pleased to announce that Hedgehog Open Network Fabric powered by SONiC is now available for download at https://githedgehog.com. We developed this release in a design partnership with our customer VMAccel. VMAccel’s distributed cloud infrastructure accelerates AI and ML inference at the data edge where low-latency cloud networks are a critical requirement.
“Many companies are looking for AI to optimize their operational efficiency. To do that, a company must balance innovation and agility against cost management. VMAccel gives our customers distributed cloud infrastructure with low-latency AI acceleration, on-prem, or in an edge environment. VMAccel tailors the accelerator cloud or micro datacenter infrastructure with Hedgehog Open Network Fabric for distributed cloud deployments of AI workloads. Hedgehog gives us a foundation of open networking and cloud network services that offers our customers a consistent, cloud-native user experience comparable to what they get with the Big 3 hyper-scalers.” -Matthew Fields, CEO of VMAccel
VMAccel’s first Hedgehog deployment is with DEMA Energy for accelerated AI edge computing infrastructure in the Kingdom of Saudi Arabia. The VMAccel infrastructure may be used for AI or ML inference for a variety of applications. The deployment requires a simple, automated, open, and low cost edge network that can be used by multiple tenants. The Infamous release of Hedgehog Open Network Fabric meets these requirements. It is powered by SONiC and certified on Dell PowerSwitch and Edgecore network devices. These switches connect GPU and FPGA servers in an immersion cooled enclosure that VMAccel can quickly deploy to edge computing locations with a fully automated network fabric that installs, boots and configures in just a few minutes. VMAccel can then use the Hedgehog API to provision fully isolated Hedgehog Virtual Private Clouds for multiple tenants to access VMAccel FPGA compute resources independent of each other. This means that multiple AI or ML applications can all use the same accelerated GPU and FPGA computing resources in an organized and secure manner.
You can download the Infamous release as a single binary that consists of the following components:
First: a Hedgehog seeder that establishes root of trust, installs SONiC on trusted devices, installs the Hedgehog control agent on trusted devices, performs UEFI secure boot of SONiC, installs the Hedgehog controller on a control node, and uses the Hedgehog controller to configure SONiC. The seeder can install, configure and run a new enterprise cloud fabric in less than 15 minutes.
Second: the Dell Enterprise SONiC network operating system that is tested, certified and supported by Hedgehog for Dell S5248F-ON switches in a collapsed core topology.
Third: a Hedgehog controller built on Kubernetes that makes it possible for a cloud infrastructure team to operate a cloud network with the same tools, processes and skills they employ for public cloud infrastructure. VMAccel uses the same Kubernetes control plane to deploy workloads and operate network and GPU/FPGA compute resources in their AI cloud infrastructure.
Fourth: a Hedgehog Virtual Private cloud service that you can use to provision multiple tenants with isolation on the same infrastructure. You can provision and manage Hedgehog VPCs with a simple Kubernetes API that should be familiar to anyone who has deployed software on an Amazon VPC.
Fifth: the option to run Hedgehog Open Network Fabric either on physical infrastructure or in a virtual lab environment. You may want to run the Hedgehog Virtual Lab to evaluate the fabric on SONiC virtual switches and Linux virtual servers. The Hedgehog support portal has instructions for downloading the software in either mode, and we have included a series of lab exercises that you can run to evaluate the software.
For more details, the Infamous release notes, and a free software download, please sign up for the Hedgehog Design Partner Program at https://githedgehog.com/design-partner-program/
For more details on VMAccel and their GPU and FPGA-accelerated AI cloud and edge solutions, visit https://vmaccel.com/