During the 2024 Open Compute Project (OCP) Global Summit, a keynote was given by GEICO outlining the reasons why in 2023 they made the decision to reverse direction for their compute infrastructure and move it back from the cloud to on-premises. Server and storage resources had become unfit, with unfit profiles for GEICO workloads while escalating costs for those resources made the cloud model impractical for their business.

The move to on-premises would carry a number of perceived challenges. The accepted model for enterprises and next-wave datacenters had been to stress the capital budget by partnering with OEMs, who would deliver more cookie-cutter solutions. The OEM hardware model turned out to be effective for managing workloads but carried limitations and little flexibility.

Figure 1 below shows the issues Geico identified with Cloud and OEM solutions.

Open-source computing was a new consideration. Widely used to outfit hyperscale datacenters and driven primarily by cloud service providers, open source was considered to be unattainable for enterprises and next-wave datacenters that didn’t carry the same volume requirements or development resources. But the benefits of being able to deliver on performance, offering simplified management and flexibility with open standard support in a future-proof lower-cost solution were too good for GEICO to pass up.

Figure 2 below shows the advantages GEICO uncovered by working through an OCP model.

Today, using OCP inspired and approved hardware for enterprise and next-wave datacenter builds is highly attainable. Reliability of configurable CRB-proven hardware has been steadily increasing, providing new areas of flexibility. One flexible area is uniformity of server management and platform security through Redfish and DC-SCM 2.0. Complimenting hardware flexibility are firmware/ software packages from suppliers like AMI that ensure interoperability and deliver on firmware updates, security and full datacenter resource management. With the introduction of these products and features, enterprise customers are able to fill the gaps that OEMs typically provide through their product uniformity.

Logistically speaking the OCP model is far more workable today as well. Participating ODMs are seeing more healthy business opportunities where product hardware is now under commoditization pressures. This is further enabled through new standards that are giving ODMs a greater ability to own product definition, where they can accept requirements from enterprise customers. And where integration and deployments have been requiring support and expertise, OCP contributors like AMI are able to provide the tools and expertise that get the datacenters up and running.

Conclusion

In the end, GEICO leveraged OCP partnerships to build out their datacenter infrastructure saying they had achieved multiple successful deployments. The presentation concluded with some challenges for consideration going forward. These challenges will continue to get answered as enterprise and next-wave datacenters look more toward contributor expertise and the available firmware, software and tools that fill the gaps for a successful deployment. Today AMI is engaging with its ecosystem partners to ensure this model works, going forward. Solutions that will address the challenges identified by GEICO will be available to better streamline implementation.