Centre de données

How

Modernized Datacenter Infrastructure

Desired Outcomes
Today’s data centers are a balancing act of capital acquisition, optimized utilization of resources, and cost control. More and more, specialized compute systems such as GPUs are required for AI and ML workloads. These systems are fairly homogeneous and have minimal network requirements. They are dynamically repurposed by the application orchestration engine so they stay fully utilized. They consume large amounts of electricity and generate massive amounts of heat. To compensate for this, the network needs to be dynamically reconfigured regularly according to the needs of the business.

Technical Requirements
Modern data centers can be cost effective and operationally simple by adhering to a set of standard principles. Equipment should be as standardized and inexpensive as possible, this maximizes the horsepower of the business, but it requires selecting non-proprietary protocols and software tools. Supply chain issues can be mitigated by selecting software that can run on a large variety of hardware platforms. This also has the benefit of providing a single software platform that needs to be automated. In addition, all systems should have a high level of compatibility, ideally a single and common API can be used to program every piece of the infrastructure.

Hedgehog’s Value
Hérisson SONiC is a fully open source network operating system that runs on a large number of platforms, eliminating supply chain issues. Hedgehog Open Network Fabric eliminates the need for proprietary vendor management software by enabling the use of the Kubernetes API, the most popular data center API, on the physical network. This empowers the DevOps team to manage the network themselves without requiring an entirely new skill set. Integration between Hedgehog Open Network Fabric and the compute environment ensures that application deployments can create and consume the necessary network services automatically without administrative intervention.