In a prior post Intel Rack Scale Design: Paving the Road to SDI and the Cloud, I gave a quick overview of Intel Rack Scale Design. This post describes how Ericsson has used Rack Scale to build a hyper scale infrastructure.
Intel Rack Scale Design Refresher
You can get more details on Intel Rack Scale Design by reading my earlier post, but here’s a quick refresher to remind you of the essentials:
- It’s a set of API specifications and open-source reference software that allows you to pool resources.
- It’s a reference architecture for a modern rack-architecture design that OEMs and infrastructure vendors like Ericsson can implement to create software-defined infrastructure (SDI) solutions for enterprise companies, telecoms, and cloud service providers.
The Intel Rack Scale Design APIs are based on and extend the DMTF Redfish™ open industry-standard specification and schema. The schema specifies a RESTful interface and utilizes JSON and OData to help customers integrate solutions within their existing data-center environments.
Data centers implementing solutions that conform to Intel Rack Scale Design can significantly improve efficiency and enable rapid service provisioning, thereby helping organizations gain cloud-like efficiency, flexibility, and agility.
Ericsson HDS 8000 Implements Intel Rack Scale Design
OEMs and ISVs are taking note of Intel Rack Scale Design and are creating solutions that take advantage of its ability to disaggregate and pool system resources. One such solution is the Ericsson Hyperscale Datacenter System (HDS) 8000. The Ericsson HDS 8000 is a hyper-converged and software-defined-infrastructure (SDI) solution based on Intel Rack Scale Design and on servers built with Intel Xeon processors and equipped with Intel network interface cards (NICs).
Ericsson HDS 8000 can transform the way that your data center works by disaggregating bare-metal compute, storage, and networking resources into virtual pools that can be flexibly reallocated to compose servers for different needs and different environments. As a result, IT workloads running on an Ericsson HDS 8000 solution can be optimized for efficiency and performance, with automated deployment and orchestration, all managed from a single pane of glass.
With the Ericsson HDS 8000, compute, storage, and networking resources are physically assembled into performance-optimized data centers (PODs). PODs are configured and controlled through management software, which enables the platform to be a software-defined, cloud-based infrastructure.
The physical structure of the Ericsson HDS 8000 makes it possible to change hardware components as technology evolves, without having to replace other interconnected components prematurely. That capability allows you to extend your infrastructure lifecycle.
The Ericsson HDS 8000 enables pooling of compute, storage, network, and graphics processing unit (GPU) nodes from PODs into virtual PODs, or vPODs. The resources of a vPOD can be physically dispersed, but because of the optical backplane in Ericsson HDS 8000, they function with the speed of a single logical system. vPODs are configured and managed using Ericsson Command Center.
You can use the Ericsson Command Center portal to assign to a vPOD whatever POD hardware is best suited to a particular workload. All storage, compute, and networking infrastructure is allocated as needed from different physical machines or racks. And because all infrastructure is software-defined, it’s simple to scale or modify a vPOD as needed to correctly size it for a given workload.
vPODs provide a number of benefits compared to traditional infrastructure, including greater agility, higher utilization, and simplified management.
More Intel Rack Scale Design Solutions
Other vendors are also creating solutions based on Intel Rack Scale Design. See, for example, American Megatrend’s “Rack Scale Design Solutions” and Supermicro’s “Supermicro Rack Scale Design (RSD) Solutions.” These solutions allow data-center operators to manage pools of storage, compute, and networking resources across racks the same way that traditional individual server and storage nodes have been managed to date.
Learn more about Intel Rack Scale Design today and decide if it’s right for yo