A comprehensive analysis of the "Red Cloud" project at Cornell University

Recently, Cornell University's top Computing center has released the "Red Cloud", a subscription based on-demand research and computing services. The project aims to enable more researchers and students to use HPC computing resources at Cornell University to conduct scientific research and discoveries to promote the discovery of scientific research. This article interprets the red cloud from three aspects, such as the services provided by the Red Cloud, the underlying configuration of the red cloud, and the benefits of using the Red Cloud.

The "Red Cloud" offers two services: the IaaS layer and the SaaS layer

The IaaS (infrastructure as a service) layer, which provides services for renting virtual servers and storage on demand, allows more researchers to use the HPC computing resources of Cornell University on Demand, based on open source cloud computing platform Eucalyptus.

▲

SaaS (Software as a service) layer, by providing "MATLAB Red Cloud" service, for Cornell University users to provide seamless connection from the private site to Mdcs worders, access to the NVIDIA GPU to optimize performance, "Matlab Red Cloud" It also provides a "fast queue" for science gateways.

The underlying configuration of the red cloud

Both of these red cloud services are running on Dell's PowerEdge C-series servers using Dell's HPC solution.

The "Red Cloud" consists of 8 servers, each configured with 2 2.7GHz of Intel Xeon E5650 CPU,48GB memory. The highest-configured virtual machines supported are 12 cores, 48GB of RAM, and virtual machines can access 7.5TB storage via 10Gb iSCSI sharing. Users subscribing to this service can use most Linux distributions, including custom CentOS 5.6 and CentOS 6, which have root permissions to access virtual servers and virtual disks. The user-selectable virtual server configuration includes the following 5 types: M1.small (1 cores, memory 4GB, disk 20GB); C1.medium (2 cores, memory 8GB, disk 40GB); M1.large (4 cores, memory 16GB, disk 200GB); M1.xlarge (8 and, memory 32GB, disk 400GB); C1.xlarge (12 cores, memory 48GB, disk 1000GB).

"The Red Cloud of MATLAB" includes 8 servers in total, each server is configured with 2 2.4GHz of Intel Xeon E5820 CPU,24GB memory, and 1 nvidia Tesla M2070 GPU, the server through Gigabit Ethernet interconnection, Mount 8TB of CIFS storage, installed software including MATLAB MDCS 2011B and CAC parallel MATLAB client code. The user subscribing to the service uses a MATLAB client program with a parallel computing toolbox on the local desktop, and then uses the MATLAB distributed computing server to extend vertically to the red cloud based on Matlab.

Eight advantages of using "Red Cloud"

Predictable, reproducible, and reliable performance. The Red Cloud service is a flexible computing and data Analysis resource for Cornell University researchers and their collaborators released by the High Computing Centre, providing predictable, reproducible, and reliable performance;

No hidden costs. Red Cloud as a subscription service, excluding the implied costs, and the first subscription to the service can also obtain 50GB of storage space, if additional storage space is required, Cornell University subscribers will be in accordance with the 0.91 U.S. dollars GB year fees, external users in accordance with 1.45 U.S. dollars GB year fees;

Fast access to data. Users can quickly access cloud internal data and data outside the cloud via 10Gb Ethernet without extra charge.

Expert support. System, application and programming consulting;

Subscription mode is easy to budget;

Compatible with the Amazon API. Easy for users to migrate from the Red Cloud platform.

These features are attractive to researchers working on data-driven applications.

▲ comparison of the advantages and disadvantages of red cloud and self-built HPC cluster

(Responsible editor: The good of the Legacy)