InCloudRack, a Rack Scale Architecture Works towards Realizing the Goal of 'Rack as a Computer'

By CIOReview | Tuesday, September 8, 2015
825
1448
281

SAN FRANCISCO, CA: In the event, Intel Developer Forum (IDF) 2015, Inspur exhibited the proof of concept for Rack Scale Architecture with the title ‘InCloudRack’. Leijun Hu, Vice President of Inspur exclaims that the Inspur InCloudRack is an ideal solution to private cloud IaaS, anticipates a better future of rack scale system and achieves the goal of "rack as a computer".

The modular design of the rack scale system consists of centralized power; cooling and control modules; along with various nodes like storage nodes, GPU coprocessor nodes and others creating the richest expansion solution across the industry. Besides the hardware expansion solution, it also incorporates hardware management, virtualization and resource scheduling software. Hardware management and resource scheduling dynamically regulates the infrastructure in accordance with operation load perceived, improving the efficiency and flexibility at the user end.

The computing, storage and network devices within the rack can be viewed as a pool of rack-level hardware resources called ‘Rack as a Computer’. The network design of InCloudRack triumphs over the traditional network with the enhanced efficiency of node communication, which decreases the communication delay by 77 percent and solves I/O problems. Currently, InCloudRack is all-set to be marketed as a ready-to-use solution.

Inspur has built two series of rack solutions in converged architecture, one being the ‘Inspur SmartRack’ favorable to public cloud customers with fast deployment and the other being ‘Inspur InCloudRack’ characterized by high availability and intelligence.

The concept of converged architecture was first proposed in 2014 focusing on Software-Defined Data Center (SDDC). Hu continues to add that the company has already realized the centralized pooling of storage and networking, is planning the next step to centralize the pooling of computing and memory which forms a complete SDDC.