Abstract

The data center hardware (HW) infrastructure in public and private clouds is going through a paradigm shift due to the demands of a heterogenous computing model. The lower-cost economics of a private cloud need to be augmented with the operational simplicity of a public cloud. The Open Compute Project (OCP) started with the goal of simplifying the compute aspect and they did succeed to some extent, but enterprise adoption has been slow due to the HW complexity of various elements and version compatibility. Since the evolution of OCP, a number of technology changes have occurred, and a new infrastructure standard is necessary to tackle the heterogenous computing model along with a thermal and high-speed input/output (I/O) interconnection nexus. Techniques are presented herein that support a converged rack architecture (CRA). Such an architecture encompasses a universal rack I/O paradigm comprising embedded opto-electrical data I/O posts, centralized power and cooling, and a common rack management unit (RMU) which allows a rack itself to be thought of as a compute unit where different compute, network, and storage components may be composed in a virtual domain on the fly and then decomposed when they are not required. In essence, a CRA manages and treats a rack frame as a complete converged system of compute, network, and storage elements and not as disparate entities.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.

Share

COinS